skill-stocktake

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local bash scripts (scan.sh, quick-diff.sh, and save-results.sh) to perform inventory tasks, identify changed files, and persist evaluation results. These scripts use standard POSIX utilities like find, jq, and awk, and include validation logic to ensure operations are confined to expected skill directories.
  • [PROMPT_INJECTION]: The skill processes markdown content from other skills which presents an indirect injection surface. 1. Ingestion points: The scan.sh script and Phase 2 auditor subagent read the contents of all skill markdown files in the global and project directories. 2. Boundary markers: The auditing process is constrained by a specific checklist and holistic judgment criteria defined in SKILL.md. 3. Capability inventory: The skill can update or delete local skill files during the consolidation phase. 4. Sanitization: All destructive operations, such as retiring or merging skills, require explicit manual confirmation from the user through a confirmation prompt before any changes take effect.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 10:29 AM
Security Audit — agent-trust-hub — skill-stocktake