workflow-skill-designer

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to perform administrative file system tasks, such as creating directories for new skill packages (mkdir -p), which is consistent with its primary purpose as a code generator.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8) because it processes untrusted data (user-provided descriptions and external requirements documents) to generate executable skill files.
  • Ingestion points: Processes free-text user descriptions and external requirements documents (via the Read tool) in phases/01-requirements-analysis.md.
  • Boundary markers: Employs markdown headers and structured templates to delimit generated content, although these do not prevent adversarial injection.
  • Capability inventory: The skill utilizes Write, Edit, Bash, Read, Glob, and Grep to manage and create skill packages.
  • Sanitization: Includes a sanitizePhaseContent function in phases/03-phase-design.md designed to strip CLI flag parsing and specific invocation syntax from generated files, but it lacks comprehensive filtering for malicious payloads within the input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 06:13 AM
Security Audit — agent-trust-hub — workflow-skill-designer