clerk-testing
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [SAFE]: The skill promotes security best practices by explicitly warning against using production keys in tests and recommending the use of test-specific API key prefixes (
pk_test_*,sk_test_*). The reference to an external repository is pinned to a specific commit hash, which is a recommended practice for integrity. - [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to fetch information from external URLs, creating an ingestion surface for untrusted data. However, the links target the official documentation and repositories of a well-known service, which mitigates the risk.
- Ingestion points: Documentation URLs for Clerk and Playwright/Cypress guides in
SKILL.md. - Boundary markers: None provided.
- Capability inventory: The skill is scoped to fetch documentation via
WebFetchand provide setup instructions. - Sanitization: None provided.
- [METADATA_POISONING]: The metadata identifies the author as 'clerk', while the skill source is attributed to a different author ('delorenj'). This is noted as a descriptive attribution to the technology provider rather than a deceptive impersonation attempt.
Audit Metadata