gathering-ui

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various shell commands to perform build verification and visual testing using local development tools.
  • Evidence: In SKILL.md, the skill executes gw dev ci --affected --fail-fast to check for compilation errors after subagent activity.
  • Evidence: In SKILL.md, the skill uses uv run --project tools/glimpse glimpse to perform showroom audits, matrix captures, and interactive visual verification.
  • [PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by passing user-controlled data into subagent prompts.
  • Ingestion points: The {ui_spec} variable used in references/conductor-dispatch.md is populated by the user's initial UI request in Phase 1 of SKILL.md.
  • Boundary markers: The dispatch templates in references/conductor-dispatch.md do not utilize boundary markers (such as XML tags) or instructions to treat the user input as untrusted data.
  • Capability inventory: The subagents dispatched with this input have the capability to create and modify source code files within the project directory.
  • Sanitization: There is no evidence of sanitization or validation performed on the user-provided UI specification before it is interpolated into the prompts for the 'Chameleon' or 'Deer' subagents.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 11:23 PM
Security Audit — agent-trust-hub — gathering-ui