4d-v21
Pass
Audited by Gen Agent Trust Hub on Jun 5, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface Risk.
- Ingestion points: The main instructions in
SKILL.mddirect the agent to check for alocal/directory and prioritize reading any markdown files within it (local/*.md) before providing developer guidance. - Boundary markers: There are no specified delimiters or "ignore embedded instructions" warnings for the content read from this directory, which might be populated with untrusted data in collaborative environments.
- Capability inventory: The skill leverages capabilities to read the file system, use search tools like
grep, and provides documentation for executing host-level shell commands via theSystemWorkerclass. - Sanitization: No sanitization or validation of the local content is implemented.
Audit Metadata