The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local automation scripts, specifically ../../scripts/check-handoff.mjs and ../../scripts/new-seed.mjs, to manage workflow state, validate handoffs between agents, and generate design seeds.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted design briefs provided by users, which are normalized and passed into the context of specialized agents (direction-scout.md and visual-critic.md). This creates an attack surface where instructions embedded in a brief could potentially influence the behavior of the delegated agents.
Ingestion points: External design briefs are ingested and normalized into ../../contracts/design-context.schema.json as specified in SKILL.md.
Boundary markers: The skill requires mandatory validation of context and handoffs using schema-based scripts (check-handoff.mjs) and prescribes the use of fresh, read-only isolated contexts for scouts and critics.
Capability inventory: The workflow utilizes file system editing, application execution, screenshot capture, and subprocess execution via Node.js scripts.
Sanitization: Untrusted brief data is validated against strict JSON schemas and handoff contracts using automated scripts before reaching downstream agents.

uiux-design