skills/simota/agent-skills/saga/Gen Agent Trust Hub

saga

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from multiple sources (customer feedback, journey maps, research findings, feature proposals) to construct stories and narratives.
  • Ingestion points: SKILL.md defines input patterns from other agents (Cast, Field, Voice, Spark, Compete, Trace), which ingest data like customer quotes and research findings.
  • Boundary markers: The skill requires human validation for BrandScript outputs and includes explicit 'Never' instructions to avoid unfounded emotional manipulation or fabricated data. The workflow includes a REFINE phase focused on anti-patterns.
  • Capability inventory: The skill generates narratives and uses the multi engine mode to delegate generation tasks to sub-agents (Codex, Antigravity, Claude).
  • Sanitization: Every generated narrative must pass the AP-1 through AP-9 anti-pattern checklist (e.g., Feature Dump, Hero Product, Jargon Wall) during the REFINE phase before delivery.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 12:16 PM