shouenlee/spec_review
plugins/full-orchestration/skills/spec_review/SKILL.md
Run a four-agent review on a technical spec or implementation plan. Use when a spec or plan needs evaluation from maintainability, security, efficiency, and completeness perspectives. Spawns four reviewer agents that write structured comment files, then merges all comments into the document in a single conflict-free pass.
testing
Updated Apr 14, 2026
$ install --global
skillsauthnpx skillsauth add shouenlee/ghcp-dev-plugin spec_reviewInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 14, 2026, 2:28 AM57.7s1 file scanned
SKILL.md
- name:
- spec_review
- description:
- >-
Spec Review
State file: .claude/swe-state/{ticket-id}.json
Four parallel reviewer agents each write a structured comment file. A single merge pass then inserts all comments into the document, avoiding parallel edit conflicts.
Usage
/spec_review .claude/specs/PROJ-123.md # Spec review (2C)
/spec_review .claude/specs/PROJ-123-impl.md # Impl plan review (2E)
Phase 1: Gather Context
Detect mode from path: {ticket-id}.md → spec mode, {ticket-id}-impl.md → impl mode. Extract ticket ID from the filename by stripping suffixes in order: -impl.md, -context.md, .md. The first match wins (e.g., PROJ-123-impl.md → strip -impl.md → PROJ-123; PROJ-123.md → strip .md → PROJ-123).
Locate state: .claude/swe-state/{ticket-id}.json. Read state to get: stages.intake.ticket_file and stages.spec.context_file.
Read: document under review (required), ticket.json (if available), context doc (if available).
Phase 2: Parallel Review
Delete any existing comment files from prior invocations before starting:
rm -f .claude/swe-state/{ticket-id}/comments-*.json
Spawn FOUR agents in a SINGLE message. Each writes to its own comment file — no direct document edits.
subagent_type: full-orchestration:{ReviewerName}
max_turns: 10
prompt: |
State file: .claude/swe-state/{ticket-id}.json
Review mode: {spec|impl}
Comment file: .claude/swe-state/{ticket-id}/comments-{ReviewerName}.json
Read state to locate the document under review:
- spec mode: stages.spec.spec_file
- impl mode: stages.spec.impl_plan_file
Also read supporting context from state:
- stages.intake.ticket_file
- stages.spec.context_file
The document may contain comments from prior review iterations
(OPEN or RESOLVED). Follow your re-review instructions for those.
Read the document and supporting files, then write your review
comments to the comment file path above.
Agents: MaintainabilityReviewer, SecurityReviewer, EfficiencyReviewer, CompletenessReviewer.
On agent failure: log warning, proceed with remaining agents.
Phase 3: Merge Comments
After all agents complete, merge their comment files into the document in a single pass. This avoids parallel edit conflicts.
- Read all 4 comment files. Skip any that don't exist (agent may have failed or found no issues).
- Read the document under review.
- Process each entry across all files:
| Action | Operation |
|---|---|
| add | Find the line containing the anchor text. Insert > **[{SEVERITY} \| {ReviewerName} \| OPEN]** {comment} after the end of that paragraph (next blank line or heading). If anchor not found, append comment at the end of the document with a note. |
| reopen | Find the blockquote line matching anchor. Change RESOLVED to OPEN. If comment field is present, replace the comment text. |
| remove | Find and delete the blockquote line matching anchor. |
- Write the updated document back.
- Delete all 4 comment files (cleanup).
Phase 4: Count and Report
Read the document and count blockquote comments matching:
> **[{SEVERITY} | {Reviewer} | OPEN]**
Report to caller:
- Count by severity: CRITICAL, HIGH, MEDIUM, LOW
- Total OPEN comments
- No verdict — the caller (spec_writer) decides convergence
Related Skills
shouenlee/task_breakdown
tools
VerifiedTrustedCommunity
Decomposes feature descriptions or GitHub issues into ordered subtasks with file-level scope, acceptance criteria, and optional gh issue create. Use when asked to "break down task", "decompose feature", "create subtasks", "plan implementation", "break this into tasks", "task list", or "implementation plan".
SKILL.mdUpdated Apr 14, 2026
shouenlee/release_notes
documentation
VerifiedTrustedCommunity
Generates changelogs from conventional commits, bumps semantic versions in pyproject.toml or package.json, and publishes GitHub releases. Use when asked to "create release", "release notes", "changelog", "bump version", "semantic version", "tag release", "publish release", or "what changed since last release".
SKILL.mdUpdated Apr 14, 2026
shouenlee/python_lint_fix
development
VerifiedTrustedCommunity
Runs ruff, mypy, and bandit on changed Python files — explains violations and auto-fixes with ruff check --fix. Use when asked to "lint", "fix lint", "check types", "type check", "run mypy", "run ruff", "python quality", or "lint python files".
SKILL.mdUpdated Apr 14, 2026
shouenlee/make_skill_template
development
VerifiedTrustedCommunity
Create new Agent Skills for GitHub Copilot from prompts or by duplicating this template. Use when asked to "create a skill", "make a new skill", "scaffold a skill", or when building specialized AI capabilities with bundled resources. Generates SKILL.md files with proper frontmatter, directory structure, and optional scripts/references/assets folders.
SKILL.mdUpdated Apr 14, 2026