Refactor Copilot Review Instructions and Skills

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

[ptr727]

Summary

• add a provider-independent code-review skill that routes changed files to the applicable language, documentation, and workflow skills
• reduce the Copilot instruction file to a review bootstrap while preserving its repository-specific disproved-claims ledger
• require visible findings and a machine-readable coverage marker
• make pr_review.py parse the marker and block unstated coverage with exit 45

Root Cause

The Copilot instruction file duplicated provider-independent rules and embedded API mechanics already implemented by scripts/pr_review.py. Copilot review output also had no stable machine-readable coverage contract, while an unstated coverage result did not produce a blocking exit code.

Validation

• uvx ruff@latest check .
• uvx ruff@latest format --check .
• uvx mypy@latest
• full coverage and self-test sequence from OPERATIONS.md (79% total coverage, 97% for scripts/pr_review.py)
• python3 scripts/build_dist.py --check
• python3 scripts/repo_gate.py
• complete prose, JSON, spec, EditorConfig, shellcheck, PSScriptAnalyzer, and Markdown gates
• 688 script tests and 45 agent-safety install tests

Follow-Up

• Add Manifest-Driven Verbatim Tree Propagation #797 distributes the review skill to downstream repository head trees for GitHub-hosted Copilot.
• Validate the structured Copilot review marker on live reviews #798 records live validation and hardening of the structured review marker.

Closes#793

[Copilot]

Pull request overview

Refactors the GitHub Copilot review bootstrap to route through a provider-independent code-review skill, and updates scripts/pr_review.py + tests to accept a stable machine-readable coverage marker and to block “unstated coverage” with a dedicated exit code.

Changes:

• Introduce a new code-review skill (in both .agents/skills and the Claude plugin distribution) that standardizes review procedure and the structured coverage marker.
• Slim .github/copilot-instructions.md down to a bootstrap + output contract while preserving the repository’s disproved-claims ledger.
• Extend scripts/pr_review.py and its tests to parse <!-- fleet-review: ... --> and to return exit 45 when coverage is unstated.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
scripts/tests/test_pr_review.py	Updates corpus/status tests to cover the new structured marker and new exit-45 behavior.
scripts/pr_review.py	Adds marker parsing and blocks unstated coverage with exit 45.
AGENTS.md	Routes “reviewing a change set” to the new code-review skill.
.github/copilot-instructions.md	Reduces to review bootstrap/runbook + preserves disproved-claims ledger.
.agents/skills/code-review/SKILL.md	Adds provider-independent review skill and defines the structured coverage marker contract.
.agents/skills/pr-review-conduct/SKILL.md	Updates “mechanics live elsewhere” guidance to point to scripts/pr_review.py + bootstrap.
.agents/skills/copilot-instructions-keeper/SKILL.md	Adjusts description to reflect bootstrap role of Copilot instructions.
.claude-plugin/fleet-skills/skills/code-review/SKILL.md	Distributes the new code-review skill into the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md	Mirrors the same “mechanics live elsewhere” update for the Claude plugin skill set.
.claude-plugin/fleet-skills/skills/copilot-instructions-keeper/SKILL.md	Mirrors the bootstrap-description update for the Claude plugin skill set.
.claude-plugin/fleet-skills/.source-digest	Updates the generated skill-set digest.
.claude-plugin/fleet-skills/.claude-plugin/plugin.json	Registers the new code-review skill in the plugin manifest.

Suppressed comments (1)

scripts/tests/test_pr_review.py:911

• The docstring for this test still says that treating a no-coverage body as a failure would “cry wolf”, but scripts/pr_review.py status now blocks on coverage=unstated with exit 45. Update the docstring to reflect that this body shape is recognized as coverage=unstated (not UNVETTED), but it is still a blocking outcome for status because it provides no proof of full diff coverage.

 def test_a_round_stating_no_coverage_at_all_is_unstated(self) -> None:
"""28 of the 332 bodies are an overview and a change list, and that shape is current.
It interleaves with the counted one throughout rather than preceding it, and one pull
request carries both across its two rounds, so failing on it would cry wolf on about one

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ptr727]

Copilot review round 4955381803 reported Suppressed comments (1).

scripts/tests/test_pr_review.py:911: The docstring still says that treating a no-coverage body as a failure would "cry wolf", while unstated coverage now blocks with exit 45.

Fixed in f1c24d7. The test now states that this is a recognized coverage=unstated shape, but it blocks because it cannot prove full diff coverage.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (2)

scripts/tests/test_pr_review.py:2844

• This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md. Reading the generated .github/skills/ file here would better validate the actual Copilot-facing contract (and would fail if the generated tree wasn’t regenerated/committed).
  .github/skills/code-review/SKILL.md:17
• The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/. In downstream repos (where .agents/skills/ is not carried), this instruction is impossible to follow and will prevent the intended routing to the per-language/workflow skills.

3. Load every applicable skill from `.agents/skills/`:

[ptr727]

Suppressed finding responses

Review round on e1396f8 reported Suppressed comments (1).

• scripts/tests/test_pr_review.py:911: "The docstring for this test still says that treating a no-coverage body as a failure would 'cry wolf', but scripts/pr_review.py status now blocks on coverage=unstated with exit 45."
  • Fixed in f1c24d7 - The fixture and its documentation state that unstated coverage is recognized but blocks the gate. A regression test asserts exit 45 for that state.

Review round on da7249f reported Suppressed comments (2).

• scripts/tests/test_pr_review.py:2844: "This contract test reads the marker from the .agents/skills/ copy, but the runbook explicitly points Copilot at .github/skills/code-review/SKILL.md."
  • Fixed in c3738e5 - The contract test reads the generated Copilot-facing skill.
• .github/skills/code-review/SKILL.md:17: "The GitHub Copilot skill lives under .github/skills/, but this step tells the reviewer to load skills from .agents/skills/."
  • Fixed in c3738e5 - The canonical skill now routes to applicable sibling skills in the current distribution. Regeneration applies that provider-neutral instruction to both generated trees.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/build_dist.py:180

• is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root (for example an accidental hand-added README.md or .DS_Store) would not be detected as drift and build_dist.py --check could still pass. Since the docs and help text describe --check as failing whenever a generated distribution differs from .agents/skills/, it should also treat any non-directory entry under the distribution root as stale (and ideally require the root directory to exist even when there are zero skills).

[ptr727]

Review round on c3738e5 reported Suppressed comments (1).

• scripts/build_dist.py:180: "is_stale() only considers directory entries when validating .github/skills/ (and the plugin's skills/), so a stray file at the distribution root ... would not be detected as drift."
  • Fixed in 04f4bf7 - Both generated skill roots must exist and contain only the expected skill directories. Three regression tests cover a stray GitHub file, a stray Claude plugin skill file, and a missing empty GitHub distribution.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.

Suppressed comments (1)

scripts/pr_review.py:19

• The status subcommand docstring says exit 0 implies “the round covering the head read the whole diff”, but report_verdict() returns 0 early when there are no head-covering reviews (if not head_reviews(pr): return 0). This makes exit 0 ambiguous (no review yet vs full coverage), so the docstring should describe the no-review case explicitly.

[ptr727]

Review round on 04f4bf7 reported Suppressed comments (1).

• scripts/pr_review.py:19: "The status subcommand docstring says exit 0 implies 'the round covering the head read the whole diff', but report_verdict() returns 0 early when there are no head-covering reviews."
  • Fixed in ced266c - The CLI contract explicitly distinguishes the no-review success case and directs callers that require review presence to wait. A contract test protects the documented behavior.

[Copilot]

Pull request overview

Copilot reviewed 53 out of 53 changed files in this pull request and generated no new comments.