ac-state-tracker

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICAL
Full Analysis
  • Indirect Prompt Injection (SAFE): The skill reads from local state files (e.g., feature_list.json, .claude/master-state.json) to manage session state. While these are untrusted data ingestion points, this behavior is essential to the skill's primary purpose and poses minimal risk in a local development context. Evidence: 1. Ingestion: feature_list.json, .claude/*.json; 2. Boundary markers: Absent; 3. Capability inventory: File writes and Git commits; 4. Sanitization: Not specified.
  • Persistence Mechanisms (SAFE): Persistence of state is the skill's primary function and is implemented using standard local file operations and Git within the project directory.
  • Data Exposure & Exfiltration (SAFE): No evidence of sensitive file access, hardcoded credentials, or unauthorized network operations.
  • Unverifiable Dependencies & Remote Code Execution (SAFE): The skill references local scripts (scripts/state_tracker.py) which were not provided, but the documented use cases are benign.
  • Prompt Injection (SAFE): No override or bypass markers were found.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 17, 2026, 06:30 PM