skill-lifecycle-manager

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill uses subprocess.run() in scripts/check.py to invoke local validation tools and the main reporting script. These commands are constructed using fixed list arguments, which prevents shell command injection. The execution is confined to internal scripts within the repository's known path structure.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes content from SKILL.md and evals.json files across the repository to classify their lifecycle stages.
  • Ingestion points: scripts/lifecycle_report.py reads and analyzes the content of files located in the skills/ subdirectories.
  • Boundary markers: No explicit delimiters are used to wrap the ingested content, though the analysis is restricted to YAML frontmatter and specific keyword regex searches.
  • Capability inventory: The skill possesses the ability to execute local Python scripts via subprocess.run, but lacks capabilities for network communication or external file writes.
  • Sanitization: The skill uses yaml.safe_load() for frontmatter parsing, protecting against unsafe object deserialization during metadata analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 06:15 AM
Security Audit — agent-trust-hub — skill-lifecycle-manager