control-tower

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration were found in the skill body or the associated rule files.
  • [COMMAND_EXECUTION]: The skill uses shell commands specifically for git worktree management (git worktree list, git worktree remove) and branch cleanup. These operations are restricted to the .claude/worktrees/ directory and are aligned with the skill's stated purpose of project maintenance.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: Analysis confirmed that the skill does not access sensitive system paths (e.g., SSH keys, cloud credentials) or perform network requests to untrusted external domains.
  • [INDIRECT_PROMPT_INJECTION]: The skill uses local project files within the spec/ directory to track workflow progress. While this involves reading untrusted data, the operation is limited to state detection and does not expose high-privilege capabilities to the ingested content.
  • Ingestion points: Files in the spec/ directory and its history/ subdirectory.
  • Boundary markers: Not explicitly defined in the state-detection logic.
  • Capability inventory: File reading, git operations, and orchestration of other internal skills.
  • Sanitization: Standard path traversal protections are assumed to be handled by the agent platform's tool constraints.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 12:17 AM
Security Audit — agent-trust-hub — control-tower