carpenter
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external 'Architect Blueprints' and human-edited documents, which are untrusted data sources that could contain malicious instructions. * Ingestion points: Processing of the blueprint in Workflow Step 1 and cataloging of human edits in Workflow Step 6 of SKILL.md. * Boundary markers: The instructions do not specify any delimiters or safety warnings to ensure the agent ignores instructions embedded within the processed data. * Capability inventory: The skill possesses file-writing capabilities (creating draft-N.md and draft-N-human-edits.md) and workflow control via AskUserQuestion, which could be targeted by an injection attack. * Sanitization: There is no evidence of input validation or content filtering for the ingested data.
- [METADATA_POISONING]: The YAML metadata in SKILL.md identifies the author as 'dmitry' via a GitHub link, which is inconsistent with the developer context that identifies the author as 'jeffallan'. This represents a minor provenance discrepancy.
Audit Metadata