skills/cedanl/.github/create-skill/Gen Agent Trust Hub

create-skill

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses Bash to execute standard developer commands such as git, gh, npx, and claude. These tools are used for repository management, searching skill registries, and automating pull request creation. This is consistent with the skill's stated purpose and developer-focused environment.
  • [EXTERNAL_DOWNLOADS]: The workflow involves searching for existing skills using the skills.sh registry and inspecting repositories from well-known organizations like Anthropic and Vercel. These operations are limited to metadata inspection and require explicit user consent for any installations, following safe development practices regarding external dependencies.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of new tools by ingesting data from user interviews, existing local skills, and external GitHub repositories via npx skills add.
  • Ingestion points: Data enters via the user interview process (Step 4) and by inspecting external registry content (Step 1).
  • Boundary markers: The agent is explicitly instructed in Step 9 to present the full SKILL.md draft and wait for user approval before proceeding to the writing phase.
  • Capability inventory: The skill uses Write, Edit, and Bash to create directories and files, and opens PRs using the gh CLI.
  • Sanitization: All generated skills must pass through a mandatory validation script validate-skill.py (Step 8) which enforces schema and safety rules before persistence. This represents a robust human-in-the-loop mitigation for indirect prompt injection risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 07:11 AM