bug-hunt
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as user-reported bug symptoms and system log output. This creates a surface where malicious instructions could be embedded in data that the agent interprets. 1. Ingestion points:
SKILL.md(Step 1) and Step 3 (Observed output/Logs). 2. Boundary markers: The skill does not define specific delimiters for untrusted content. 3. Capability inventory: The skill allows shell command execution (pnpm vitest,grep,git bisect) and local file writing (.bug-hunts/). 4. Sanitization: The instructions explicitly mandate redacting sensitive information like tokens and keys from logs before saving. - [DYNAMIC_EXECUTION]: The agent is instructed to perform instrumentation by inserting temporary probes into the codebase and creating reproduction scripts, involving runtime generation and execution of code to verify hypotheses.
- [COMMAND_EXECUTION]: The workflow relies on executing shell-based commands for testing and searching, such as
pnpm,vitest,grep, andgitto facilitate the debugging process.
Audit Metadata