second-brain-reflection
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill extracts 'rules' and 'stratagems' from missions, PRs, and experiments, which are then written to persistent files like
~/.gemini/GEMINI.md. If the source task data contains malicious instructions, they could be persisted across sessions and influence future agent behavior. - Ingestion points: Task history, PR descriptions, and experiment results processed during the 'Reflection Protocol'.
- Boundary markers: No explicit markers are defined to isolate untrusted task content from the reflection and storage logic.
- Capability inventory: The skill instructs the agent to write to file system paths including
./GEMINI.mdand the~/.gemini/directory. - Sanitization: No sanitization or validation of extracted knowledge is performed prior to file persistence.
- [NO_CODE]: The skill consists solely of markdown instructions and YAML metadata without any accompanying scripts, binaries, or source code files.
Audit Metadata