cosmic-mythic-overlay
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data about space objects, which represents a potential surface for indirect prompt injection.
- Ingestion points: The
objectinput parameter defined inSKILL.mdandmanifest.jsonaccepts arbitrary string data. - Boundary markers: The workflow includes a 'Separation pass' and 'Lock the astronomy' step to conceptually distinguish between factual data and mythic framing.
- Capability inventory: The skill is restricted to using the
Readtool and lacks access to network operations, file system writing, or command execution. - Sanitization: The skill uses a structured JSON output contract to enforce data separation but does not specify explicit input sanitization or filtering.
Audit Metadata