apply-clean-code

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external code for review and refactoring, which introduces a surface for indirect prompt injection where malicious instructions could be embedded in the code provided for analysis.\n
  • Ingestion points: Code snippets or files provided for review or refactoring as defined in SKILL.md and agents/claude.md.\n
  • Boundary markers: The instructions do not specify the use of delimiters or explicit warnings to ignore instructions embedded within the code being analyzed.\n
  • Capability inventory: The agent has the capability to perform filesystem operations and execute the provided Python scripts (scripts/init_benchmark.py and scripts/summarize_benchmark.py) along with standard agent tools.\n
  • Sanitization: No explicit sanitization, escaping, or validation of the input code is defined in the instructions or scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 03:53 AM
Security Audit — agent-trust-hub — apply-clean-code