retro
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes session logs, which are external and dynamic data sources. Malicious instructions or adversarial content embedded within these logs (e.g., in error messages or code output) could influence the agent's analysis or suggestions during the retrospective process.
- Ingestion points: Local session logs retrieved from the machine (SKILL.md, Step 2).
- Boundary markers: Absent; the skill does not specify the use of delimiters or instructions to ignore embedded commands within logs.
- Capability inventory: The skill searches through session logs, reads repository configuration files (e.g., package.json), and suggests or applies changes to steering files (CLAUDE.md, AGENTS.md) and coding standards.
- Sanitization: No sanitization or validation of the log content is described.
- [COMMAND_EXECUTION]: The skill instructs the agent to identify and execute existing repository check commands (e.g., linting or testing scripts) to verify their effectiveness.
- Evidence: The instructions state, "Read the repo's own check command first... and prove it goes red on a real past mistake."
Audit Metadata