workflow-review-changes
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute local Node.js utility scripts located within the project's hidden directory to manage workflow contracts and review snapshots.
- Evidence:
node .claude/scripts/lib/workflow-skill-contract.cjs --writeinSKILL.md. - Evidence:
node .claude/hooks/lib/review-receipt.cjs snapshot --target=<worktree|staged|commit-descriptor|branch>inreferences/fix-loop.md. - Evidence:
node .claude/hooks/lib/review-receipt.cjs issue --kind=workflow-review-changes --scope=full-changeset --snapshot-json='<exact saved JSON>'inreferences/fix-loop.md. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (pull requests and code changes) and possesses significant capabilities, creating a surface for potential injection attacks.
- Ingestion points: The skill reads pull request changes, local project files, and the workflow registry (
.claude/workflows.json) as part of its triage and review process. - Boundary markers: The provided instructions do not specify explicit delimiters or "ignore instructions" warnings to wrap the code being reviewed.
- Capability inventory: The skill utilizes
spawn_agentfor subagent coordination, executes localnodescripts, and has full read/write access to project reports and temporary workflow data. - Sanitization: There is no evidence of sanitization or escaping of the code content before it is processed by the agent or passed to subagents.
Audit Metadata