detecting-doc-drift
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [SAFE]: No malicious patterns, such as prompt injection, data exfiltration, or obfuscation, were detected. The skill instructions maintain a professional and technical focus on documentation consistency.
- [COMMAND_EXECUTION]: The skill utilizes specific commands involving the
doclifecyclePython module (e.g.,python3 -m doclifecycle drift-plan). These operations are consistent with the skill's primary purpose of analyzing a repository and are executed within the local environment. - [PROMPT_INJECTION]: The skill ingests untrusted content from documentation and code files, presenting a potential surface for indirect prompt injection. 1. Ingestion points: Various files within the audited repository, including
README.md,CLAUDE.md, and source code. 2. Boundary markers: The instructions do not specify explicit delimiters for the text being analyzed. 3. Capability inventory: Execution of thedoclifecyclemodule and a validation script. 4. Sanitization: The skill includes a dedicated script,validate-drift-output.py, which enforces a strict JSON schema for the audit results and specifically filters for shell-sensitive characters in evidence citations to prevent the model from being coerced into generating executable payloads.
Audit Metadata