skills/veged/caveman/caveman-commit/Gen Agent Trust Hub

caveman-commit

Pass

Audited by Gen Agent Trust Hub on Apr 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's primary function is restricted to generating text based on provided diffs. It does not interact with the file system or execute commands.- [COMMAND_EXECUTION]: The instructions explicitly state that the skill must not run 'git commit', stage files, or perform any other shell-based operations, mitigating the risk of unauthorized command execution.- [PROMPT_INJECTION]: The skill provides clear formatting rules and does not contain any instructions aimed at bypassing AI safety protocols or extracting system information.- [DATA_EXFILTRATION]: There are no network-capable commands (e.g., curl, wget) or external URL references that could be used for data transmission.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of git diffs.
  • Ingestion points: Reads git diffs provided by the user or agent environment.
  • Boundary markers: None explicitly defined for diff content.
  • Capability inventory: Limited to text output in a code block; no file-write, network, or execution capabilities.
  • Sanitization: None specified for input content. Despite the ingestion of untrusted data, the restricted capability set ensures this surface cannot be used for malicious impact.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 17, 2026, 10:35 AM
Security Audit — agent-trust-hub — caveman-commit