audit-dead-code
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from a repository's codebase. Maliciously crafted instructions hidden in code comments, documentation, or string literals (e.g., within tags like
TODO,deprecated, orunusedwhich the skill specifically targets) could potentially influence the agent's behavior during the audit or the auto-fix process. - Ingestion points: The agent reads the entire codebase, including source files, configuration files, manifests, and documentation.
- Boundary markers: The instructions do not specify the use of delimiters or clear boundaries to isolate the audited code content from the agent's internal instructions.
- Capability inventory: The skill utilizes file system read capabilities for searching and mapping, and file system write capabilities for the 'Auto-fix' functionality.
- Sanitization: While the instructions emphasize building a 'proof chain' and 'guarding against false positives' through manual verification steps, there is no automated sanitization or filtering of the ingested content to prevent instructions in the code from being interpreted as commands by the LLM.
Audit Metadata