bug-killer
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFE
Full Analysis
- [DYNAMIC_EXECUTION]: The skill generates and executes reproduction scripts, tests, and diagnostic probes to confirm root causes. This is a core function of the debugging process and involves human-in-the-loop confirmation before repair implementation.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data such as logs, stack traces, and repository files. While it uses a structured 'Chain of Evidence' format to track investigation, there is a theoretical risk of malicious instructions within ingested data influencing the agent's behavior.
- Ingestion points: SKILL.md, user bug reports, system logs, and repository source code.
- Boundary markers: Uses structured Markdown headers and a defined case template to separate data fields.
- Capability inventory: File writing to the '/coe/' directory and shell command execution for diagnostics and reproduction.
- Sanitization: Relies on the agent's internal reasoning and user confirmation gates; no explicit programmatic sanitization is defined.
Audit Metadata