adding-a-skill
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The instructions include a validation command referencing an absolute path to a Python script in a specific user's directory:
/Users/steve/.codex/skills/.system/skill-creator/scripts/quick_validate.py. This is non-portable and references internal file structures on a developer's machine. - [PROMPT_INJECTION]: The skill defines a process for creating and updating skill definitions (
SKILL.md). Because the agent is instructed to read and validate these files, it creates a surface for indirect prompt injection where malicious instructions inside a skill file could attempt to influence the agent's behavior during the validation or deployment steps. Evidence: - Ingestion points:
skills/<skill-name>/SKILL.md,README.md(via Plain Public Skill Checklist) - Boundary markers: None provided in instructions
- Capability inventory:
python3,node,npm,pnpm(via validation and smoke-test steps) - Sanitization: None mentioned for processing the target skill files
- [EXTERNAL_DOWNLOADS]: The skill references the use of
npx @agent-native/skills@latestfor installation. This is a vendor-owned package associated with the BuilderIO organization.
Audit Metadata