duck-debt
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, unauthorized data exfiltration, or dangerous command executions were detected. The skill operates in a read-only capacity.
- [PROMPT_INJECTION]: A heuristic flag for instruction override was found to be a false positive. The skill mentions patterns like 'ignore prior rules' specifically as examples of malicious content to be ignored within repository data, rather than as an instruction to the agent itself.
- [INDIRECT_PROMPT_INJECTION]: The skill proactively addresses the risks associated with processing untrusted data from repository comments.
- Ingestion points: Repository files and comments scanned during the 'Scan for signals' phase in SKILL.md.
- Boundary markers: Explicit instructions in the 'Untrusted text handling' section command the agent to treat repo content as data and never as instructions.
- Capability inventory: The skill is limited to read-only operations; it contains no file-write, network-send, or code-execution capabilities.
- Sanitization: Matched snippets are restricted to 160 characters and stripped of control characters and markdown delimiters before being reported.
Audit Metadata