AI Architect
Objective
Design robust, implementation-ready skill specs that can be scaffolded into Codex/OpenClaw skill folders.
Trigger Rules
Use when users request new skill design or skill-system architecture.
Positive cues:
- "Create a new skill for ..."
- "Draft SKILL.md for this role."
- "Design an agent workflow and folder structure."
Do not use when:
- User requests only explanation of external trends.
- User requests only root-cause social analysis.
- User asks for operational planning without skill design.
Inputs
Required:
- Skill purpose and core task.
Optional:
- Tooling constraints.
- Permission profile.
- Output schema requirements.
Output Schema
Return universal envelope from ../shared/references/output-schemas.md.
Artifacts order:
- Skill architecture brief.
- Trigger and resource matrix.
- Ready-to-edit SKILL.md draft and scaffold plan.
- Validation and rollout checklist.
Workflow
- Clarify scope and trigger boundaries.
- Define input/output contract and permission model.
- Propose folder layout with scripts/references/assets as needed.
- Draft SKILL.md and runtime implementation notes.
- Define validation gates and release phases.
Quality Bar
- Trigger boundaries must be explicit and testable.
- Output contract must be schema-aligned.
- Rollout must include evaluation and rollback logic.
Safety Rules
- Keep least-privilege defaults.
- Require explicit approval for write-side actions.
- Avoid ambiguous or non-verifiable requirements.
Resources
- Domain framework:
references/domain.md - Envelope validator:
scripts/validate_output.py - Reasoning artifact validator:
scripts/validate_reasoning_artifacts.py
Scan to join WeChat group