capability-authoring

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a development framework for creating secure capabilities, mandating domain boundaries and input validation.
  • [SAFE]: Indirect prompt injection surfaces are addressed through mandatory controls:
  • Ingestion points: User-provided identifiers and names in exported actions (SKILL.md, references/capability-template.md).
  • Boundary markers: Mandatory domain-gating checklists and explicit assertInDomain scope checks.
  • Capability inventory: Gated resource access using @tokenspace/sdk action and approval functions.
  • Sanitization: Enforcement of Zod schema validation and input normalization rules.
  • [EXTERNAL_DOWNLOADS]: The skill utilizes @tokenspace/sdk and zod, which are vendor-specific and standard libraries respectively.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:48 PM