lov-checkpoint

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local Python utility scripts, scripts/profile_store.py and scripts/validate_skill.py, to handle profile data and verify the skill source. These scripts are invoked locally by the agent to maintain user-owned configuration and ensure deployment quality.
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes external and potentially untrusted data sources including Git history, diffs, and documentation. Ingestion points: .checkpoint_log, Git status/history, README.md, and CHANGELOG.md. Capability inventory: Writing to the local filesystem and staging Git commits. Boundary markers: The execution boundary defines a strict separation between analysis/preview and actual modification, requiring explicit scope for write actions. Sanitization: The skill implements manual review boundaries and separates preview from execution to mitigate instruction injection risks.
  • [SAFE]: The profile storage utility (profile_store.py) includes specific security controls such as a blacklist for sensitive keys (e.g., 'password', 'token', 'api_key') to prevent accidental exposure of secrets. It also manages files in standard local directories using atomic writes and restrictive file permissions (0o600).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 09:47 AM
Security Audit — agent-trust-hub — lov-checkpoint