vapa-exec

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs the agent to preserve raw issue snapshots and save comments and quoted acceptance criteria verbatim (which would cause any secrets present in those sources to be copied into workspace artifacts), creating a risk of the LLM handling/outputting secret values even though it later advises redaction as a mitigation.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). vapa-exec’s Step 1 “Fetch the Source of Truth” fetches and ingests the raw GitHub issue body and all comments (saved as issue.json and individual files under comments/), which are outsider-authored free text.