Skip to content

Fix ASVS 5.0 numbering across PRD skill, data index, and examples - #7

Merged
Xcaciv merged 3 commits into
mainfrom
claude/plugin-security-review-sijami
Aug 25, 2026
Merged

Fix ASVS 5.0 numbering across PRD skill, data index, and examples#7
Xcaciv merged 3 commits into
mainfrom
claude/plugin-security-review-sijami

Conversation

@Xcaciv

@Xcaciv Xcaciv commented Aug 25, 2026

Copy link
Copy Markdown
Collaborator

The bundled data/asvs files are genuine ASVS 5.0 (V6 Authentication,
V8 Authorization, V16 Logging), but the chapter index in data/asvs/README.md,
the PRD skill's chapter mappings, its coverage-gap table, its worked example,
and the example enhanced PRD all used pre-5.0 numbering (V2 Authentication,
V4 Access Control, V12 Input Validation) or cited requirements that do not
exist (V2.5, V11.1.4). Every coverage matrix the skill produced therefore
cited requirement IDs that resolve to unrelated requirements.

  • Rebuild the chapter index from the data files' own title frontmatter
  • Correct all mappings and gap-table citations to verified 5.0 IDs
  • Mark above-level items (e.g. 6.3.8 is L3) as recommended escalations
  • Add numbering-discipline instructions to the skill
  • Add scripts/check_refs.py validating every V/S reference in instruction
    surfaces against the bundled data (negative-tested)
  • Add eval assertions failing the PRD workspace on pre-5.0 chapter labels

Co-Authored-By: Claude Fable 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_017LupEJCBDKT1xRyzNHiW4q

claude added 2 commits August 24, 2026 22:19
The bundled data/asvs files are genuine ASVS 5.0 (V6 Authentication,
V8 Authorization, V16 Logging), but the chapter index in data/asvs/README.md,
the PRD skill's chapter mappings, its coverage-gap table, its worked example,
and the example enhanced PRD all used pre-5.0 numbering (V2 Authentication,
V4 Access Control, V12 Input Validation) or cited requirements that do not
exist (V2.5, V11.1.4). Every coverage matrix the skill produced therefore
cited requirement IDs that resolve to unrelated requirements.

- Rebuild the chapter index from the data files' own title frontmatter
- Correct all mappings and gap-table citations to verified 5.0 IDs
- Mark above-level items (e.g. 6.3.8 is L3) as recommended escalations
- Add numbering-discipline instructions to the skill
- Add scripts/check_refs.py validating every V/S reference in instruction
  surfaces against the bundled data (negative-tested)
- Add eval assertions failing the PRD workspace on pre-5.0 chapter labels

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017LupEJCBDKT1xRyzNHiW4q
Packaging fixes (verified against current Claude Code plugin docs):
- Move slash commands from .claude/commands/ (project-only, never scanned
  for installed plugins) to commands/ at the plugin root, rewritten as thin
  dispatchers so each skill remains the single authoritative procedure
- Promote the FIASSE lookup command to a proper skill (skills/fiasse-lookup)
  so natural-language FIASSE/SSEM questions can trigger it
- Add a path-resolution preamble to every skill: bundled data resolves via
  CLAUDE_PLUGIN_ROOT in plugin installs, relative to the skill file elsewhere
- Stamp release versions into .claude-plugin/plugin.json (the manifest Claude
  Code actually reads) and exclude dev-only dirs from the release zip
- Correct .claude/settings.json permission rule syntax

Portability (AGENTS.md / Agent Skills standards):
- Add AGENTS.md as the canonical tool-agnostic entry point; CLAUDE.md now
  imports it (plugin installs never loaded CLAUDE.md; README claim corrected)
- Add scripts/install_skills.sh: layout-preserving installer for opencode
  (.opencode, .claude, .agents, or global config), keeping skills' internal
  references working unchanged
- Add CI workflow: reference checker, JSON/manifest validation, shell syntax
  checks, plugin layout verification

Skill hardening:
- Generation and review skills now run existing project linters/tests/
  scanners and dependency audits as evidence when present, and say so when
  absent instead of implying verification
- Review skill treats reviewed code as data, never instructions

Add docs/critical-review-2026-08.md: critical assessment of whether the
plugin increases securability in practice, with a prioritized enhancement
plan. Bump version to 2.1.0.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017LupEJCBDKT1xRyzNHiW4q
Copilot AI lite review requested due to automatic review settings August 25, 2026 02:39

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR corrects ASVS 5.0 chapter/requirement numbering across the PRD enhancement skill, ASVS data index, and examples, and adds automated validation to prevent future reference drift across instruction surfaces.

Changes:

  • Fix ASVS chapter mappings and example outputs to consistently use ASVS 5.0 numbering (e.g., V6 Authentication, V8 Authorization, V16 Logging).
  • Add scripts/check_refs.py plus CI/release workflow steps to validate ASVS/FIASSE references against bundled data/.
  • Improve plugin portability/packaging (path-resolution guidance, installer script, plugin zip version stamping, and root-level commands/ dispatchers).

Reviewed changes

Copilot reviewed 26 out of 26 changed files in this pull request and generated 1 comment.

Show a summary per file
File Description
tests/prd-securability-enhancement-workspace/evals/evals.json Reformats eval JSON and adds an assertion enforcing ASVS 5.0 chapter-label discipline.
skills/securability-engineering/SKILL.md Adds explicit path-resolution guidance and encourages evidence via running existing project tooling.
skills/securability-engineering-review/SKILL.md Adds path-resolution guidance, anti-instruction-steering note, and tool-evidence guidance during reviews.
skills/prd-securability-enhancement/SKILL.md Updates ASVS 5.0 mappings, adds numbering-discipline instructions, and corrects gap-table/example ASVS references.
skills/fiasse-lookup/SKILL.md Promotes the lookup content into a proper skill with frontmatter and clarified query handling.
scripts/install_skills.sh Adds a layout-preserving installer for agent tools outside Claude Code plugin installs.
scripts/check_refs.py Adds a reference validator for ASVS/FIASSE citations across instruction surfaces.
scripts/build_plugin_zip.sh Excludes dev-only content from release artifacts and stamps version into .claude-plugin/plugin.json.
README.md Updates installation guidance, structure, and documents the new lookup capability and plugin layout.
examples/prd-enhancement/enhanced-prd.md Updates the example enhanced PRD to ASVS 5.0 chapter numbering.
docs/critical-review-2026-08.md Adds a critical review doc explaining defects found (incl. ASVS numbering drift) and fixes/plan.
data/asvs/README.md Rebuilds the chapter index to match bundled ASVS 5.0 files and warns about renumbering vs 4.x.
commands/secure-generate.md Adds root-level command dispatcher delegating to securability-engineering skill.
commands/securability-review.md Adds root-level command dispatcher delegating to securability-engineering-review skill.
commands/prd-securability-enhance.md Adds root-level command dispatcher delegating to prd-securability-enhancement skill.
commands/fiasse-lookup.md Adds root-level command dispatcher delegating to fiasse-lookup skill.
CLAUDE.md Converts to a thin stub importing AGENTS.md (canonical guidance).
AGENTS.md Adds canonical, tool-agnostic entry point content and repo layout/guidance.
.github/workflows/release.yml Runs scripts/check_refs.py as part of release validation.
.github/workflows/ci.yml Adds CI workflow for reference validation, JSON parse checks, shell syntax checks, and layout verification.
.claude/settings.json Fixes Claude Code permission pattern syntax to Read(path/**) form.
.claude/commands/secure-generate.md Removes old project-only command file (now replaced by root commands/).
.claude/commands/securability-review.md Removes old project-only command file (now replaced by root commands/).
.claude/commands/prd-securability-enhance.md Removes old project-only command file (now replaced by root commands/).
.claude-plugin/plugin.json Bumps plugin version to 2.1.0.
.claude-plugin/marketplace.json Bumps marketplace manifest version to 2.1.0.

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

Comment thread skills/prd-securability-enhancement/SKILL.md Outdated
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
@Xcaciv
Xcaciv merged commit 76d2168 into main Aug 25, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants