maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project source code and documentation which are external and potentially untrusted, creating a surface for indirect prompt injection attacks where malicious code or comments could influence the agent's behavior.
- Ingestion points: In Step 2 ("Source wave"), the skill launches subagents to read feature files from the project source to identify documentation drift.
- Boundary markers: The skill does not define explicit boundary markers or delimiters for the ingested code, though it implements structural safeguards by restricting subagents to read-only roles ("Children never drive the app and never edit files").
- Capability inventory: The skill possesses the ability to spawn subagents (
spawn_agent), drive the application for live verification (Step 4), and write corrections to the local verification skill's files (SKILL.md, features/, and harness scripts) in Step 6. - Sanitization: The instructions do not mention explicit sanitization, escaping, or filtering of the ingested content before it is processed or incorporated into generated recipes and PRs.
Audit Metadata