faq
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill processes external data from repositories (issues, tickets, PRs, chat transcripts) which could contain malicious instructions. The skill mitigates this by explicitly stating that such content must be treated as data and never as instructions.
- Dynamic Context and Command Execution: The skill involves identifying repository test commands (e.g.,
npm test,pytest) and interacting with tools. It includes a specific safeguard against command injection by instructing the agent to never put values returned from automated scouting into shell commands. - Consent and Human-in-the-Loop: For potentially sensitive operations like writing a local file (
CLAUDE-FAQ.md) or creating calendar events, the skill mandates explicit user confirmation and human review of the content before execution. - Data Privacy: There are explicit instructions to never copy secrets or credential-file contents and to avoid reading configuration text directly, relying instead on user-provided information for defaults.
Audit Metadata