caveman-compress

Pass

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python module provided within its directory structure.
  • Evidence: It instructs the agent to run the command python3 -m scripts <absolute_filepath> to perform the compression logic.
  • [COMMAND_EXECUTION]: The skill performs extensive file system operations including reading, creating backups, and overwriting files.
  • Evidence: Instructions specify that the compressed version overwrites the original file while a human-readable backup is saved as FILE.original.md.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from local memory files, which presents a surface for indirect prompt injection.
  • Ingestion points: Reads and processes content from files such as CLAUDE.md, todos, and preferences.
  • Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" warnings for the data being processed.
  • Capability inventory: The skill can overwrite files and execute local scripts via subprocesses.
  • Sanitization: No sanitization, escaping, or validation of the ingested file content is specified.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 24, 2026, 09:05 AM
Security Audit — agent-trust-hub — caveman-compress