change-impact
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external data, such as code changes and consumer implementations, which introduces an indirect prompt injection surface where malicious instructions embedded in analyzed files could attempt to influence the agent's behavior.\n
- Ingestion points: The agent is directed to inspect committed changes, staged/unstaged edits, and new files within the repository (SKILL.md).\n
- Boundary markers: The instructions do not specify the use of delimiters or instructions to ignore embedded commands to isolate analyzed code from the agent's control flow.\n
- Capability inventory: The agent is expected to read repository files, trace references, and potentially observe runtime behavior through tests or execution.\n
- Sanitization: The methodology does not include requirements for sanitizing or filtering the content of the analyzed files for malicious instructions.\n- [NO_CODE]: The skill consists exclusively of markdown documentation and methodology guidance; it does not include any scripts, configuration files for package managers, or executable binaries.
Audit Metadata