diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a rigorous debugging discipline and instructions for the agent to follow. No malicious patterns or behaviors were detected.
- [DATA_EXPOSURE]: The skill includes strong guidelines for data safety, explicitly instructing the agent to redact secrets and credentials from command outputs and captured artifacts before they are displayed or processed.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data such as network traces, log dumps, and codebase content. While this creates an attack surface for indirect prompt injection, the risk is inherent to the primary function of debugging and is mitigated by the instructions' focus on specific error symptoms and structured feedback loops.
- Ingestion points: Processes external files including GLOSSARY.md, ADRs, network payloads, HAR files, and log dumps.
- Boundary markers: Instructions require redacting secrets and capturing only relevant signal lines.
- Capability inventory: Involves running shell commands, test runners, and instrumentation via the agent's available tools.
- Sanitization: Explicit redaction policy for sensitive information.
Audit Metadata