deepspace
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches Markdown documentation, index files (llms.txt), and the full documentation corpus (llms-full.txt) from the official vendor-managed domain docs.deep.space. This allows the agent to access authoritative and up-to-date SDK information during execution.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon external documentation content from the vendor's site. This creates a surface where instructions within the documentation could influence agent behavior.
- Ingestion points: Retrieval of llms-full.txt and individual guide files from docs.deep.space.
- Boundary markers: No specific delimiters or 'ignore' warnings are mandated for the external documentation content.
- Capability inventory: Project scaffolding, deployment, and secret management via the deepspace CLI tool.
- Sanitization: Documentation is processed as raw text to maintain technical accuracy.- [SAFE]: The skill includes explicit security guidance, such as instructing the agent to use proper secret management tools (deepspace secrets), avoid putting sensitive data in command line arguments or logs, and rely on verified JWTs for caller identity rather than internal client-controlled headers.
Audit Metadata