agent-creator

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python script via the Bash tool to perform automated validation of agent files.
  • Evidence: python3 ${CLAUDE_SKILL_DIR}/../skill-creator/scripts/validate-skill.py --agents-only {plugin-dir}/ located in SKILL.md.
  • [EXTERNAL_DOWNLOADS]: The documentation references and recommends reading official specifications from Anthropic's documentation site.
  • Evidence: Reference to https://code.claude.com/docs/en/sub-agents in references/anthropic-agent-spec.md. This is a well-known service domain.
  • [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface because it reads and parses existing agent files for validation.
  • Evidence: The 'Validation Workflow' in SKILL.md instructs the agent to 'Locate the agent .md file' and 'Parse YAML frontmatter'.
  • Ingestion points: Existing agent files located via Glob/Read.
  • Boundary markers: Standard YAML and Markdown structural markers.
  • Capability inventory: Read, Write, Edit, Glob, Grep, Bash, AskUserQuestion.
  • Sanitization: The instructions focus on structural and schema validation rather than executing content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 11:30 AM
Security Audit — agent-trust-hub — agent-creator