socratic-agentic-workflowlisted

# Socratic agentic workflow skill This skill guides the LLM through a structured interview with the user to define the high-level logic of an agentic system — before any code is written. The output is a coherent set of linked MD documents that capture intent, structure, and rationale, ready for handover to a coding agent or human builder. **Target session length: 30 minutes.** **Do not write code. Stay at the logic and intent layer.** Implementation context — where the system runs, what tools or platforms it must use, how credentials are handled — is still in scope. That's what theme 6 (Constraints) is for. The line to hold is "no code", not "no implementation". --- ## How this skill works The skill runs in three phases: 1. **Interview** — adaptive, Socratic questioning across 8 fixed themes plus drift detection 2. **Synthesis** — AI reflects on answers, flags contradictions, surfaces assumptions, and runs an agent design pass that checks every agent role against five quality conventions 3. **Output** — two silent pre-generation passes ground the scaffold and mission in current reality (using `WebSearch` when relevant), then generate a named, linked set of MD files — including an index file — the user can take away Read all linked documents before starting: - [`INTERVIEW.md`](./INTERVIEW.md) — question themes, pacing rules, and drift detection logic - [`SYNTHESIS.md`](./SYNTHESIS.md) — how to reflect, challenge, and confirm before generating output