brainstorming
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process vague feature ideas, ambiguous requirements, and large text dumps from the user, which could contain malicious instructions meant to influence agent behavior.
- Ingestion points: Requirements are gathered in
SKILL.md(Step 1: Assess and ground) andreferences/interview-and-approaches.md(Phase 1: Understand the Idea), including substantial initial context dumps. - Boundary markers: The workflow implements explicit confirmation gates: a "scoping synthesis" checkpoint in
SKILL.md(Step 5) andreferences/scope-synthesis.md, and a final "interactive approval" gate inSKILL.md(Step 7). - Capability inventory: The skill has the capability to write design documents to the local filesystem (
docs/brainstorms/) and invoke interactive tools likeAskUserQuestion(described inSKILL.md). - Sanitization: The skill relies on the LLM's synthesis of requirements and mandatory user confirmation of interpreted scope to mitigate accidental obedience to embedded instructions.
Audit Metadata