explain-pr
Pass
Audited by Gen Agent Trust Hub on Aug 5, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains explicit defensive instructions to protect against indirect prompt injection. It instructs the agent to treat all external PR content (descriptions, commits, diffs) strictly as data for explanation and never as instructions. Ingestion points: The agent reads PR metadata and diffs using git and PR viewing commands (SKILL.md). Boundary markers: The skill uses explicit instructions to treat data as material to explain and mandates a rigid output structure. Capability inventory: The skill uses shell-based tools for repository inspection. Sanitization: Includes a mandatory secret redaction process.
- [CREDENTIALS_UNSAFE]: A clear protocol for secret redaction is included. The skill mandates scanning all code snippets for API keys, tokens, and passwords, replacing them with a redacted placeholder to prevent accidental exposure of credentials.
- [DATA_EXFILTRATION]: The skill minimizes exfiltration risk by constraining all output to the user chat. It explicitly prohibits the agent from posting comments or edits to external GitHub or GitLab instances.
- [COMMAND_EXECUTION]: The skill utilizes standard, well-known CLI tools for data retrieval. These commands are focused on local state inspection and do not execute untrusted external scripts.
Audit Metadata