retro
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes session transcripts, tool outputs, and repository steering files, which creates a potential surface for indirect prompt injection where malicious data could influence agent behavior.
- Ingestion points: The agent reads session history, transcript data, and various repository files (e.g., AGENTS.md, package.json, CI workflows) as specified in the Sources section of SKILL.md.
- Boundary markers: The instructions do not define specific delimiters or guardrails to isolate ingested external data from the agent's core instructions.
- Capability inventory: The skill is instructed to propose modifications to repository configuration, steering files, linters, and automation hooks, although it explicitly mandates user review and approval before any change is implemented.
- Sanitization: There is no mention of sanitizing or validating the content ingested from session history or tool outputs before it is processed by the agent.
Audit Metadata