vibe-coding-self-improve
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from skill files and external URLs to generate proposed modifications to the agent's own instruction set.
- Ingestion points: The agent reads existing skill definitions from
skills/[skill-name]/SKILL.md(Reflection Protocol Step 2) and validates externalGitHub URLs(Autoresearch Criteria Item 5). - Boundary markers: The process includes a mandatory human-in-the-loop checkpoint in Step 4, which requires the user to explicitly approve a diff before any file writes occur.
- Capability inventory: The skill utilizes file read capabilities and a file-writing 'Edit tool' (Step 5) to apply instruction updates to the local filesystem.
- Sanitization: The skill relies on the surgical nature of the proposed diffs and manual user review to sanitize the output before it is committed to the skill repository.
Audit Metadata