find-bugs
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to execute shell commands including 'rm', 'mkdir', and 'jq', as well as development tools like 'cargo check', 'tsc', 'eslint', and 'mypy' on the target directory provided by the user.
- [PROMPT_INJECTION]: The skill uses incentivized roleplay instructions (e.g., '+10 points for critical bugs', 'high score to beat is 85') to influence the agent's behavior during the Search, Adversary, and Judge phases.
- [PROMPT_INJECTION]: Vulnerable to Indirect Prompt Injection as it ingests and analyzes untrusted source code. \n
- Ingestion points: Source files within the target directory specified by the user command /find-bugs. \n
- Boundary markers: Absent. The skill does not use delimiters to isolate untrusted code from its own system instructions. \n
- Capability inventory: Execution of system commands (rm, mkdir, jq) and language-specific linters/compilers. \n
- Sanitization: Absent. There is no evidence of code sanitization or filtering before the analysis occurs.
Audit Metadata