workflow
Fail
Audited by Snyk on Nov 17, 2026
Risk Level: CRITICAL
Full Analysis
CRITICAL E004: Prompt injection detected in skill instructions.
- Potential prompt injection detected (medium risk: 0.30). The visible content shows no malicious intent, but the skill silently runs an unpinned, externally published npm package (
npx codebase-intelligence) that always tracks the latest release and suppresses its errors from the user, which is a sketchy, undisclosed-to-user behavior rather than an attack.
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (low risk: 0.10). The required workflow reads GitHub-hosted content from the external codebase-intelligence repo only as first-party docs (llms.txt/llms-full.txt, trusted vendor documentation), but the codebase-intelligence and review flows also consume the user's own repo files, git diffs, and agent config; no step ingests outsider-authored free text such as issues, PRs, emails, or public feeds on a forced-delivery path, so the only outsider-reachable exposure is arbitrary web search/fetch that the agent must actively initiate (review production mode's WebSearch overlay on language best-practice pages), which is low.
MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).
- Potentially malicious external URL detected (medium risk: 0.30). The inbound dependency is the codebase-intelligence doc fetch in references/codebase-intelligence.md, which the agent is instructed to pull via WebFetch or curl from an individual-owned GitHub
mainbranch and use for command syntax and tool selection, with no digest or signature; that owner is unofficial, so the tier is medium.
Issues (3)
E004
CRITICALPrompt injection detected in skill instructions.
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
W012
MEDIUMUnverifiable external dependency detected (runtime URL that controls agent).
Audit Metadata