wiki-ingest
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill interacts with untrusted external content via
web_fetch, creating a surface for indirect prompt injection. However, the instructions include significant mitigations, such as mandatoryfetch-guardexecution and an identity validation step to prevent source spoofing. - [DATA_EXFILTRATION]: A mandatory 'Sensitive content guard' is implemented (Step 5) to scan both source content and generated prose for live credentials, API keys, and tokens. The skill is instructed to redact these findings or stop execution entirely if a secret is found in a raw source, preventing accidental credential exposure.
- [COMMAND_EXECUTION]: The skill uses the
skillwikiCLI to perform operations. The logic includes 'fail-closed' conditions where the agent must stop and report errors if guards fail or if the publication tool is unavailable, ensuring the agent does not proceed in an insecure state.
Audit Metadata