grilling-decisions
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data from project management and design tools, which creates an attack surface for indirect prompt injection.
- Ingestion points: The skill explicitly reads data from Linear tickets (descriptions, comments), Figma nodes, and other project documentation using various tools (Linear MCP, Figma MCP, Grep, Glob).
- Boundary markers: The instructions mandate quoting sources verbatim within specific markers („…“) and citing sources in parentheses, which provides some structural delimitation but does not semantically neutralize malicious instructions embedded in the data.
- Capability inventory: The skill itself focuses on decision-making, but it is a core component of a broader implementation pipeline (
lt-dev:ticket-cycle) that includes automated code writes and network operations via other skills. - Sanitization: The skill requires a "Re-check" phase to identify contradictions and evidence, which serves as a logical validation step, though it lacks explicit technical sanitization of the ingested content.
Audit Metadata