vibe-coding-state
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from a local file and user input, which presents an attack surface for indirect prompt injection.
- Ingestion points: Reads
progress.txtacross several operations (Read State, Resume, Recover) and accepts manual user input to rebuild the state during recovery. - Boundary markers: The skill defines a structured schema for
progress.txtusing---delimiters to separate blocks, which helps in parsing but doesn't prevent injection within fields. - Capability inventory: The skill has capabilities to read and append to
progress.txt, but it is restricted from modifying other files. - Sanitization: The skill includes a validation phase that checks for specific allowed values (e.g., phase names like IDEATE, DOCUMENT, etc.) and valid state transitions, which provides some mitigation against malformed or unexpected data.
Audit Metadata