observe-before-editing

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructs the agent to read and act upon data from external sources, specifically logs and cache directories, which creates an attack surface for Indirect Prompt Injection.
  • Ingestion points: The agent is directed to read directory contents (ls) and log files (tail) from .claude/cache/ and ~/.claude/ (SKILL.md).
  • Boundary markers: No specific boundary markers or instructions to ignore embedded commands within the logs are provided.
  • Capability inventory: The agent possesses capabilities to list directories, read files, run arbitrary failing commands, and edit code (SKILL.md).
  • Sanitization: There is no evidence of sanitization or validation of the content read from logs or the directory structure before the agent processes it.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 09:03 AM
Security Audit — agent-trust-hub — observe-before-editing