codekb-extract
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as source code comments, git history, and pull request descriptions which could contain malicious instructions designed to influence the extraction LLM.\n
- Ingestion points: Ingests code content and git metadata through the
codekbCLI tool across various scopes likesrc/payment.\n - Boundary markers: No specific delimiters or "ignore previous instructions" guards are defined to separate the code content from the extraction prompts.\n
- Capability inventory: Triggers execution of the local
codekbCLI tool; no direct file-write or network-exfiltration capabilities are explicitly requested in the skill manifest.\n - Sanitization: Employs a confidence-based review system where extractions with
confidence < 0.7are flagged for human review, providing a degree of protection against adversarial inputs.\n- [NO_CODE]: The skill consists solely of a markdown instruction file and does not include any accompanying scripts or executables.
Audit Metadata