start-here
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to scan external-facing project files such as the README and open issues. These files could potentially contain malicious instructions intended to influence the agent's summary or downstream actions.
- Ingestion points: README, package manifests (package.json, pyproject.toml, etc.), commit subjects, PR titles, and open issues.
- Boundary markers: The instructions explicitly tell the agent to treat the repo as "input to critique, not gospel" and to challenge existing framing.
- Capability inventory: The skill is authorized to read files and write the
founder-brief.mdfile to the filesystem. - Sanitization: No explicit programmatic sanitization is defined, but the agent is guided to maintain a conversation-driven, human-prose approach which reduces the likelihood of executing embedded commands.
Audit Metadata