capture-triage
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE] (SAFE): The submission contains only documentation files (
GUIDE.md,ROADMAP.md). No functional code, shell scripts, or YAML configuration files defining the AI skill's logic were found. - [SAFE] (SAFE): The documentation describes a 'Preview Pattern' as a core security feature, ensuring that the agent cannot route tasks or spawn external agents without explicit user approval of the classification table.
- [PROMPT_INJECTION] (LOW): As a triage tool, this skill inherently processes untrusted data (user captures).
- Ingestion points: Captures read from 'Drafts Pro' or an 'Inbox' folder.
- Boundary markers: None specified in documentation.
- Capability inventory: Moves text to a 'Ready' queue (file-write) and can trigger 'research-swarm' (agent execution).
- Sanitization: No technical sanitization mentioned, though the 'Preview Pattern' serves as a manual sanitization gate.
Audit Metadata