Restore copilot assignee resolution by preferring issue-scoped assignee checks

[Copilot]

A regression switched agent assignee validation to the repo-scoped assignee check, which does not reliably surface coding-agent bots (including copilot-swe-agent). This change restores effective resolution by validating aliases against the issue-scoped assignee endpoint when issue/PR context is available, with safe fallback behavior.

• What changed

  • Added a shared alias validator in assign_agent_helpers.cjs that:
    • first calls GET /repos/{owner}/{repo}/issues/{issue_number}/assignees/{assignee} when an issue/PR number is known
    • falls back to rest.issues.checkUserCanBeAssigned(...) when issue-scoped validation is unavailable or non-applicable
  • Updated findAgent(...) and getAvailableAgentLogins(...) to accept optional issue/PR context and route through the new validator.

• Call-site updates

  • Threaded issue/PR number into agent lookup from:
    • assign_to_agent.cjs
    • assign_copilot_to_created_issues.cjs
    • create_issue.cjs
    • create_pull_request.cjs

• Behavioral impact

  • Preserves prior successful bot assignment behavior in issue/PR flows.
  • Retains compatibility via repository-scoped fallback where issue-scoped checks cannot determine assignability.

// New validation path (simplified)awaitgithub.request("GET /repos/{owner}/{repo}/issues/{issue_number}/assignees/{assignee}",{ owner, repo, issue_number, assignee });// fallback:awaitgithub.rest.issues.checkUserCanBeAssigned({ owner, repo, assignee });

---

pr-sous-chef: requested branch update for run https://github.com/github/gh-aw/actions/runs/28133822633

Generated by 👨‍🍳 PR Sous Chef · 36.5 AIC · ⌖ 1.02 AIC · ⊞ 17.4K · ◷

[pelikhan]

@copilot Add logging to allow easier debugging

[Copilot]

Pull request overview

This PR fixes a regression in coding-agent assignee validation by preferring the issue-scoped assignee check endpoint when issue/PR context exists, with a repository-scoped fallback for compatibility.

Changes:

• Added a shared assignee-alias validator and threaded optional issue/PR context into findAgent(...) / getAvailableAgentLogins(...).
• Updated multiple call sites to pass issue numbers into agent lookup (issue creation, PR creation fallback, and assign-to-agent).
• Extended tests to cover issue-scoped validation and fallback behavior.

Show a summary per file

File	Description
actions/setup/js/assign_agent_helpers.cjs	Adds validateAssigneeAlias(...) and updates agent lookup/availability helpers to prefer issue-scoped assignability checks.
actions/setup/js/assign_agent_helpers.test.cjs	Adds test coverage for issue-scoped alias validation and fallback behavior.
actions/setup/js/assign_to_agent.cjs	Threads issue/PR number into findAgent(...) for assign-to-agent flow.
actions/setup/js/assign_copilot_to_created_issues.cjs	Passes issue number into findAgent(...) for created-issues assignment flow.
actions/setup/js/assign_copilot_to_created_issues.test.cjs	Updates expectations for findAgent(...) call signature change (now includes issue number).
actions/setup/js/create_issue.cjs	Passes issue number into findAgent(...) when auto-assigning Copilot during issue creation.
actions/setup/js/create_pull_request.cjs	Passes fallback issue number into findAgent(...) for PR flow fallback issue assignment.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 7/7 changed files
• Comments generated: 5

[Copilot]

@copilot Add logging to allow easier debugging

Added targeted debug logging in assign_agent_helpers.cjs to make alias validation flow easier to diagnose (issue-scoped check attempts, fallback to repo-scoped checks, and unexpected status failures). Included in commit 7c133ed.

[github-actions]

✅ Test Quality Sentinel completed test quality analysis.

[github-actions]

🧠 Matt Pocock Skills Reviewer has completed the skills-based review. ✅

[github-actions]

✅ Design Decision Gate 🏗️ completed the design decision gate check.

No ADR enforcement needed: PR #41306 does not have the 'implementation' label (has_implementation_label=false) and has 0 new lines of code in business logic directories (well under the 100-line threshold). The 7 changed files are skill scripts, generated .lock.yml workflows, and a shared markdown doc.

[github-actions]

✅ PR Code Quality Reviewer completed the code quality review.

[github-actions]

Skills-Based Review 🧠

Applied /diagnose and /tdd — commenting with observations; no blocking issues found.

📋 Key Themes & Highlights

Key Themes

• Missing regression test: The specific failing scenario (issue-scoped 204 + repo-scoped 404 for copilot-swe-agent) is not captured as a test case — a future revert to repo-scoped would go undetected.
• Test brittleness: The new getAvailableAgentLogins tests rely on call-order mocks that assume a specific alias ordering from AGENT_LOGIN_NAMES; mockImplementation with parameter matching would be safer.
• Undocumented fallback intent: The 404 path in validateAssigneeAlias falls back to the repo-scoped check despite 404 meaning "not assignable to this issue" in the GitHub API; a comment explaining the deliberate treatment as "inconclusive" would help maintainability.
• Parameter insertion risk: Adding issueNumber between agentName and githubClient is a positional breaking change — all call sites are updated here, but an options bag would guard against future silently-broken callers.

Positive Highlights

• ✅ Clean extraction of validateAssigneeAlias — single responsibility, good debug logging throughout
• ✅ Tight error discrimination: only 404/422 trigger fallback; all other statuses re-throw, preserving observability
• ✅ issueNumber || pullNumber in assign_to_agent.cjs elegantly handles both issue and PR contexts in one line
• ✅ All five call sites correctly updated and the existing test suite adapted to match the new call routing

🧠 Reviewed using Matt Pocock's skills by Matt Pocock Skills Reviewer · 108 AIC · ⌖ 8.35 AIC · ⊞ 6.5K

Comments that could not be inline-anchored

actions/setup/js/assign_agent_helpers.test.cjs:244

[/diagnose] The regression scenario is not directly tested: copilot-swe-agent succeeds the issue-scoped check but fails the repo-scoped check — the root cause described in the PR.

The current test has copilot-swe-agent get a 404 from issue-scoped and fall back, while github-copilot-enterprise succeeds via issue-scoped. That validates the routing logic but misses the specific regression.

<details>
<summary>💡 Suggested regression test</summary>

it(&quot;shouldfindcopilot-swe-agen…</details><details><summary>actions/setup/js/assign_agent_helpers.test.cjs:127</summary>**[/tdd]**Themocksequence`mockRejectedValueOnce→mockResolvedValueOnce→mockRejectedValue`isposition-dependent:itassumes`github-copilot-enterprise`isalwaysthe2ndaliasiteratedin`AGENT_LOGIN_NAMES`.Ifthealiasorderchanges,thetestsilentlyteststhewrongalias.
Thesameissueappearsatline146inthe422fallbacktest(`checkUserCanBeAssigned.mockRejectedValueOnce(...).mockResolvedValueOnce(...)`).&lt;details&gt;&lt;summary&gt;💡SuggestedfixusingmockImplementation&lt;/summary&gt;…
</details><details><summary>actions/setup/js/assign_agent_helpers.cjs:125</summary>
**[/diagnose]** The 404 fallback warrants a comment: per GitHub&#39;s API docs, 404 from `GET /repos/{owner}/{repo}/issues/{issue_number}/assignees/{assignee}` means &quot;user cannot be assigned to this issue&quot; — not &quot;endpoint unavailable&quot;. Falling back to the repo-scoped check on a definitive 404 could produce a false positive if the repo-scoped check returns 204.
In practice this is benign for bot logins (repo-scoped also returns 404 for them), but the intent is non-obvious and the next maintainer ma…
</details><details><summary>actions/setup/js/assign_agent_helpers.cjs:190</summary>**[/zoom-out]**Inserting`issueNumber`asanewpositionalparameterbetween`agentName`and`githubClient`isabreakingAPIchangemitigatedherebyupdatingallknowncallsites.Anycalleroutsidethisdiffthatpassedacustom`githubClient`asthe4thargwouldsilentlyreceive`issueNumber=&lt;githubClientobject&gt;`(truthy)andattempttouseitasanissuenumber.
Thesameconcernappliesto `getAvailableAgentLogins` atline82.&lt;details&gt;&lt;summary&gt;💡Consideranoptionsbagforfuture…</details>

[pelikhan]

@copilot run or-finisher skill

[github-actions]

REQUEST_CHANGES — Two high-severity bugs plus two broken tests block this merge.

Blocking issues summary

High

1. TypeError + NaN in validateAssigneeAlias (line 139, assign_agent_helpers.cjs, existing comment): if the provided githubClient lacks a .request method, a TypeError is thrown with no status property — the condition status !== 404 && status !== 422 evaluates to true, so the error is re-thrown and the intended fallback to checkUserCanBeAssigned is silently skipped. Additionally, Number(issueNumber) produces NaN for non-numeric strings and for object values, corrupting the API call without any validation guard.

2. getAvailableAgentLogins called with wrong arity in assign_to_agent.cjs (line 408, existing comment at line 335): getAvailableAgentLogins(effectiveOwner, effectiveRepo, githubClient) passes the custom client as the issueNumber parameter. This causes if (issueNumber) to be truthy (an object), Number(githubClient) = NaN is sent as issue_number, and githubClient defaults to the global github object — silently dropping the intended custom client in every error-enrichment path.

Medium

3. Broken test: getAvailableAgentLogins with issue number (line 127, assign_agent_helpers.test.cjs, existing comment): checkUserCanBeAssigned is not mocked to reject in this test. After vi.clearAllMocks(), the mock returns undefined (not a rejected promise), so await checkUserCanBeAssigned(...) resolves — all four fallback logins pass validation. The test expects ["github-copilot-enterprise"] but will receive all five aliases.

4. Broken test: findAgent with issue number (line 247, assign_agent_helpers.test.cjs, existing comment): same root cause — the first alias falls back to an unmocked checkUserCanBeAssigned which resolves, so findAgent returns on the first alias without ever reaching the second. The test asserts two request calls but only one occurs.

5. 404 fallback silently overrides "not assignable" verdict (line 125, assign_agent_helpers.cjs, new comment): status !== 404 && status !== 422 routes both statuses to the repo-scoped fallback. HTTP 404 from this endpoint means the user cannot be assigned; treating it like 422 ("not applicable") allows a rejected login to resurface as available if the repo-scoped check is more permissive, deferring the real failure to assignment time.

🔎 Code quality review by PR Code Quality Reviewer · 116.9 AIC · ⌖ 7.64 AIC · ⊞ 5.2K

[github-actions]

🧪 Test Quality Sentinel Report

⚠️Test Quality Score: 78/100 — Acceptable

Analyzed 13 test(s): 13 design, 0 implementation, 0 guideline violations. All new and modified tests verify the new issue-scoped assignee-resolution behavior as behavioral contracts.

📊 Metrics & Test Classification (13 tests analyzed)

Metric	Value
New/modified tests analyzed	13
✅ Design tests (behavioral contracts)	13 (100%)
⚠️ Implementation tests (low value)	0 (0%)
Tests with error/edge cases	8 (62%)
Duplicate test clusters	0
Test inflation detected	Yes — assign_copilot_to_created_issues.test.cjs (8 additions vs 1 production addition)
🚨 Coding-guideline violations	0

Test	File	Classification	Issues Detected
should use issue-scoped assignee checks when issue number is provided(new)	assign_agent_helpers.test.cjs	✅ Design	—
should fall back to repository-scoped checks when issue-scoped check returns 422(new)	assign_agent_helpers.test.cjs	✅ Design	See note
should prefer issue-scoped assignee checks when issue number is provided(new)	assign_agent_helpers.test.cjs	✅ Design	—
should successfully assign copilot agent(modified)	assign_agent_helpers.test.cjs	✅ Design	—
should return error when agent is not available(modified)	assign_agent_helpers.test.cjs	✅ Design	—
should report already assigned when agent is in assignees(modified)	assign_agent_helpers.test.cjs	✅ Design	—
should skip assignment when a secondary copilot alias is already assigned(modified)	assign_agent_helpers.test.cjs	✅ Design	See note
should successfully assign copilot to single issue(modified)	assign_copilot_to_created_issues.test.cjs	✅ Design	—
should handle multiple issues(modified)	assign_copilot_to_created_issues.test.cjs	✅ Design	—
should handle agent not found(modified)	assign_copilot_to_created_issues.test.cjs	✅ Design	—
should handle already assigned agent(modified)	assign_copilot_to_created_issues.test.cjs	✅ Design	—
should handle failed assignment(modified)	assign_copilot_to_created_issues.test.cjs	✅ Design	—
should handle error during assignment(modified)	assign_copilot_to_created_issues.test.cjs	✅ Design	—

JavaScript: 13 (*.test.cjs). Score breakdown: behavioral coverage 40/40 · error/edge coverage 18/30 · duplication 20/20 · proportional growth 0/10.

⚠️ Minor Observations (2)

should fall back to repository-scoped checks when issue-scoped check returns 422 (assign_agent_helpers.test.cjs) — ✅ Design, minor observation: all three assertions are call-count checks (toHaveBeenCalledTimes). Unlike test 1 and test 3, there is no URL/parameter assertion verifying which fallback endpoint is called during the 422-triggered repo-scoped retry. Suggested improvement: add toHaveBeenCalledWith('GET /repos/{owner}/{repo}/assignees/{assignee}', ...) to pin the fallback endpoint contract.

should skip assignment when a secondary copilot alias is already assigned (assign_agent_helpers.test.cjs) — ✅ Design, minor observation: the assertion was changed from expect(mockGithub.request).not.toHaveBeenCalled() to toHaveBeenCalledTimes(1). This is correct, but the test no longer pins what that single call is. Suggested improvement: add a toHaveBeenCalledWith(...) assertion to lock in the expected URL/params.

Test inflation note: assign_copilot_to_created_issues.test.cjs shows 8 additions vs 1 production addition (8:1). This is a structural artifact — 6 tests each contain an inline eval string with one findAgent call site requiring the same parameter addition — not genuine test bloat.

Verdict

✅ Check passed. 0% implementation tests (threshold: 30%). All new tests verify behavioral contracts for the issue-scoped assignee resolution feature, including the 404 and 422 error fallback paths.

🧪 Test quality analysis by Test Quality Sentinel · 182 AIC · ⌖ 17.3 AIC · ⊞ 8.4K · ◷

[github-actions]

✅ Test Quality Sentinel: 78/100. Test quality is acceptable — 0% of new tests are implementation tests (threshold: 30%).

[Copilot]

@copilot run or-finisher skill

Ran the PR finisher pass and addressed the actionable blockers in commit 4082923 (assignee validation hardening, corrected issue/client argument threading, and test fixes). Local make fmt, make lint, make test-unit, make test, and make recompile all pass.

[github-actions]

@copilot review all comments and address unresolved review feedback.
Please recheck after the latest fixes and rerun CI.

Generated by 👨‍🍳 PR Sous Chef · 36.5 AIC · ⌖ 1.02 AIC · ⊞ 17.4K · ◷

[pelikhan]

@copilot use core.info for logging
Fix https://github.com/github/gh-aw/actions/runs/28134020610/job/83316949311

[Copilot]

@copilot use core.info for logging
Fix https://github.com/github/gh-aw/actions/runs/28134020610/job/83316949311

Updated the new assignee-validation diagnostics to use core.info and fixed the failing typecheck from run 28134020610 in commit b476386.

[pelikhan]

@copilot review and fix https://github.com/github/gh-aw/actions/runs/28135181839/job/83320305623?pr=41306