narrative-registry
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill resolves its installation root using
git rev-parse --show-topleveland invokes a local Python utility script namedregistry-events.pyto handle event streaming and narrative updates. - [INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion point for brand narratives, positioning data, and user-provided examples. This represents a potential surface for indirect instructions, although the instructions specify that internal references must be validated and drafts should be treated as untrusted proposals.
- [DATA_EXPOSURE]: The skill maintains a versioned narrative stream in
memory/events/narrative.ndjson. It includes explicit safeguards to ensure host capability values are never recorded in request JSON, files, or logs, minimizing the risk of credential leakage.
Audit Metadata