publish-evidence
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external test evidence and artifacts to generate pull request comments, which represents an indirect injection attack surface.
- Ingestion points: Data is ingested from test run directories, test names, and claim evidence provided through the
--test-runparameter. - Boundary markers: The skill uses a specific
<!-- test-evidence -->comment marker to identify and manage its output in pull requests. - Capability inventory: The agent uses the
pnpm evals:e2ecommand to perform publishing operations. - Sanitization: No explicit sanitization of test names or evidence claims is described in the prompt instructions.
- [CREDENTIALS_UNSAFE]: The skill references a sensitive credential
BLOB_READ_WRITE_TOKENbut adheres to secure practices by instructing the agent to load it from the environment or a secret management service (Infisical) rather than hardcoding it.
Audit Metadata