systematic-debugging

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION] (LOW): Indirect Prompt Injection Surface. The skill instructs the agent to process untrusted data such as error messages and stack traces (Phase 1).\n
  • Ingestion points: Error messages, stack traces, and git diff output mentioned in SKILL.md.\n
  • Boundary markers: Absent.\n
  • Capability inventory: The skill utilizes shell command execution (env, security, codesign) for diagnostic purposes.\n
  • Sanitization: Absent.\n- [DATA_EXFILTRATION] (LOW): Potential Data Exposure. The skill suggests logging environment variables for debugging in Phase 1, Step 4.\n
  • Evidence: 'env | grep IDENTITY' in SKILL.md.\n
  • Risk: Logging environment variables can inadvertently expose sensitive secrets, keys, or tokens to standard output or log files if the variables match the filter pattern.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:48 PM