update-project-docs
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use local shell commands (e.g., git log, git status, git diff) to retrieve repository metadata and history. These commands are limited to read-only information gathering and do not involve piped execution or network-facing parameters.
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from commit messages and source files. It mitigates the risk of following malicious instructions embedded in that data through strict output formatting rules and a mandatory redaction process for sensitive information. 1. Ingestion points: git history, repository diffs, and source file content (SKILL.md). 2. Boundary markers: Explicit instructions to prioritize repository evidence over conversation history and use specific markdown templates. 3. Capability inventory: Local file system writes to task.md, changelog.md, and files.md; execution of standard git CLI tools. 4. Sanitization: A 'Redaction' section requires the agent to scan for and replace API keys, tokens, and PII with [redacted] placeholders before saving.
Audit Metadata