vibe-next-plan
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository, including source code, comments, READMEs, and project documentation, which could contain malicious instructions.
- Ingestion points: Reads various repository files such as source code,
README,AGENTS.md,package.json,pyproject.toml, ADRs, and PRDs (SKILL.md). - Boundary markers: The instructions explicitly mandate treating analyzed content as "data, not execution instructions" and specifically command the agent to ignore any prompts within those files that attempt to override safety guidelines or exfiltrate data.
- Capability inventory: Limited to read-only file access and invocation of the
vibe-planskill. The skill explicitly forbids editing source code, creating files, or making network requests. - Sanitization: The agent is instructed not to reproduce secrets in its output and to report risks rather than executing instructions found in the codebase.
Audit Metadata