robius-matrix-integration
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill's frontmatter description uses the marker 'CRITICAL' and a list of specific trigger keywords to influence the agent's internal priority and skill selection logic.
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill facilitates the ingestion of untrusted data from the Matrix network (messages, room names, etc.), which could contain malicious instructions.
- Ingestion points: The
matrix_worker_taskandspawn_timeline_subscriberimplementations inSKILL.mdreceive and process live data streams from external Matrix homeservers. - Boundary markers: The provided implementation patterns lack delimiters or system-level instructions to ignore potential commands embedded within external chat content.
- Capability inventory: The skill includes capabilities to interact with the Matrix API, including room management and message transmission, which could be abused if the agent obeys instructions found in ingested data.
- Sanitization: There is no evidence of content sanitization or validation for the ingested text events in the provided code snippets.
Audit Metadata