explain
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions explicitly direct the agent to "Respect project-specific instructions over global assumptions" found in files like
AGENTS.md,.codex/, and READMEs. This creates an attack surface where a malicious repository could include instructions designed to override agent behavior or provide deceptive explanations.\n - Ingestion points: Step 1 of the workflow in
SKILL.mdidentifiesAGENTS.md,.codex/,.agents/, and README files as primary sources of information and instruction.\n - Boundary markers: No instructions are provided to treat project-level guidance as untrusted or to isolate it from the skill's primary instructions.\n
- Capability inventory: The skill is limited to read-only inspection (file reading and directory mapping using
rg). It lacks network exfiltration or local file-writing capabilities unless explicitly approved by the user.\n - Sanitization: The skill does not implement any filtering or sanitization of the content read from the project files.
Audit Metadata