skills/dafang/codestable/cs-refactor/Gen Agent Trust Hub

cs-refactor

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local utility script located at .codestable/tools/codestable-worktree-gate.py using python3 to manage worktree states and validation gates during the refactoring process.
  • [PROMPT_INJECTION]: The skill processes untrusted data from the user's codebase during the 'scan' phase to generate optimization checklists. This creates a surface for indirect prompt injection (Category 8) where malicious instructions embedded in code comments could attempt to influence agent behavior in later stages.
  • Ingestion points: Project source files within the scan range identified in SKILL.md (e.g., specific components or modules).
  • Boundary markers: Absent. The instructions do not define specific delimiters or "ignore instructions" warnings for the code content being scanned.
  • Capability inventory: The agent has access to shell commands (python3, git, grep), file system write access for documentation and checklists, and the ability to modify project code during the 'apply' phase.
  • Sanitization: Absent. Scanned code content is analyzed directly without evidence of filtering or sanitization before interpolation into the refactoring logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 12:38 PM
Security Audit — agent-trust-hub — cs-refactor