observe-before-editing

Pass

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: LOW
Full Analysis
  • [COMMAND_EXECUTION] (INFO): The skill instructs the agent to use standard observation commands (ls, tail) to verify system state. These are benign and limited to the local filesystem for debugging purposes.
  • [DATA_EXFILTRATION] (SAFE): While the skill accesses local cache and log paths (e.g., .claude/cache/), it does not contain any network operations or mechanisms to transmit data externally.
  • [INDIRECT_PROMPT_INJECTION] (LOW): The skill directs the agent to read log files (tail .claude/cache/*.log). If the application being debugged logs unsanitized external input, this could theoretically serve as an ingestion point for indirect prompt injection. However, as a set of debugging guidelines for an agent's internal environment, the risk is minimal and inherent to the task of log analysis.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 15, 2026, 11:39 PM