skill-prototype

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes runtime inputs, such as a 'question' and 'hypothesis', to generate and store code or documentation artifacts. This workflow creates an entry point for potentially malicious instructions hidden in the user-provided data to influence agent behavior during file creation or cleanup.
  • Ingestion points: Processes user-provided 'question', 'hypothesis', and 'artifact_path' variables.
  • Boundary markers: The instructions lack explicit delimitation or 'ignore embedded instructions' directives for user-provided strings when interpolating them into outputs.
  • Capability inventory: The skill has the capability to create and modify files and invoke the 'scripts/plan-storage.sh' shell script.
  • Sanitization: No input validation or sanitization mechanisms are described for the incoming text content.
  • [COMMAND_EXECUTION]: The skill references the use of a local shell script for artifact management.
  • Evidence: Mentions invoking 'scripts/plan-storage.sh' to store results outside the plugin cache.
  • [EXTERNAL_DOWNLOADS]: The documentation references an external code repository as its origin and source material.
  • Reference: Cites 'mattpocock/skills' at commit '3cca18b368ae95cdbdebbff572ccafa662551015'.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 04:32 PM
Security Audit — agent-trust-hub — skill-prototype