weekly-report
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFEPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including website content, CSV exports, and pasted summaries. It contains explicit instructions in
references/agent-security.mdto treat this content as data rather than instructions, mitigating the risk of indirect prompt injection attacks. - [CREDENTIALS_UNSAFE]: The skill's instructions acknowledge that user inputs often contain secrets. It mandates the detection and redaction of credentials (API keys, tokens) and provides a protocol for notifying the user about necessary rotations without reproducing the secrets in output or files.
- [PERSISTENCE]: The skill manages persistence by reading from and writing to the
.agents/product-context.mdfile to store project-specific context. Security rules are in place to prevent the persistence of raw personal data and credentials, and the skill is required to inform the user where output is being written. - [SAFE]: The skill follows established best practices for data handling, accessibility, and user interaction. The instructions in the
references/directory serve as a security guardrail for the agent's behavior.
Audit Metadata