design-auto
Skill0 starsChanged 4 months agoMITView on GitHub
Autonomous (non-interactive) design for the autopilot. Produces docs/{task}/design.md like /design, but resolves decisions itself via a panel + pre-mortem critic instead of a human conversation. The design itself is produced by a pinned fable/max sub-agent; the orchestrator spawns, verifies, escalates. Invoked by /autopilot; never run interactively.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- dlaumets/agent-design-packtheir own version, first seen Jul 14, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- dlaumets/agent-design-pack★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- add-to-claudeSkill
- add-to-rulesSkill
- audit-claudeSkill
- audit-rulesSkill
- autopilotSkill
- better-planSkill
- designSkill
- execute-autoSkill
- executeSkill
- facts-extSkill
- factsSkill
- questionsSkill
- research-autoSkill
- researchSkill
- review-autoSkill
- structure-autoSkill
- structureSkill
- wiki-captureSkill
- wiki-ingestSkill
- wiki-lintSkill
- wiki-querySkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 30
Before declaring the design complete, verify and fix: (internal) no contradictory decisions, no "Still figuring out" item that actually …
- Distinctive: line 45
Before proposing structure or a design, CONSULT the captured, enforced system model:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
You are running design in autonomous mode for the autopilot — no human to talk to. The artifact is the same as /design (docs/{task}/design.md), in the same format, with every decision resolved autonomously and an adversarial check standing in for the human. Design is the pipeline's highest-leverage point — a design flaw survives the review gates and forces expensive rework — so the design itself is produced by a sub-agent pinned to the strongest model, not in this session.
Orchestrator: spawn the pinned design agent
Spawn ONE design sub-agent (model: fable, effort: max, no subagent_type, no isolation) with this prompt:
Read .claude/skills/design-auto/SKILL.md and follow it from "## Inputs (read first)" onward — those sections are your instructions; you are the design agent for {task}. Write docs/{task}/design.md exactly as specified (or docs/{task}/ESCALATE.md if an escalation trigger fires). Do not commit. Your final message is one line: DESIGN WRITTEN, ESCALATED, or FAILED: <reason>.When it returns: if docs/{task}/ESCALATE.md exists, handle it per the autopilot escalation flow. Otherwise verify docs/{task}/design.md exists and matches the base skill's sections (## Where we are through ## What we decided). Missing/malformed artifact or FAILED -> spawn the agent once more; a second failure -> write docs/{task}/ESCALATE.md yourself and escalate. Do NOT write or patch the design in this session — the pin exists so the strongest model does the designing.
Inputs (read first)
docs/{task}/prior artifacts (facts.md,questions.md,research.mdif present).docs/objective.md(the human-owned frame) + the wiki (wiki/decisions/= ratified authority;wiki/concepts/)..claude/skills/design/SKILL.mddefines the artifact format — match it exactly.
Autonomous decision procedure — panel + pre-mortem critic
- Generate 3 independent design approaches from deliberately different angles — e.g. (a) simplest-thing-that-works, (b) risk-first (minimize the worst failure), (c) cheapest-to-extend. Each a short candidate approach + its key decisions.
- Pre-mortem critic — for each candidate, assume it shipped and failed; name the most likely failure cause. Kill/down-weight candidates whose failure is severe + likely.
- Synthesize the survivor into the design, grafting the strongest ideas from the runners-up. Record the choice + rejected alternatives under
## What we decided. - Genuinely-unresolved items that block structure →
## Still figuring out, unless they trip an escalation trigger.
Consistency check (before the artifact is done)
Before declaring the design complete, verify and fix: (internal) no contradictory decisions, no
"Still figuring out" item that actually blocks structure, no conflict with docs/objective.md /
wiki/decisions/ / prior artifacts; (against the codebase) Glob/Grep that every claim about
existing code holds (greenfield → nothing it assumes exists conflicts). An inconsistency you can't
resolve (contradicts a ratified decision / needs the human) → docs/{task}/ESCALATE.md.
Guardrails (every mode)
- Ground in reality, never invent — every decision traces to the objective, a wiki decision, a prior artifact, or actual code/sources.
- Defaults, recorded — where
/designwould ask the human, pick the best-supported default and record it as an autonomous (provisional) decision: what, why, alternative. - Escalate instead of guessing when it matters — if a choice would contradict a human-ratified
wiki/decisions/entry, change the objective frame, or hinges on info you cannot obtain → STOP and writedocs/{task}/ESCALATE.md(the question + why it's not autonomously resolvable). Don't write a half-artifact. - Self-contained output at
docs/{task}/design.md, matching the base skill's sections.
Architecture & design constraints (stewardship — read before deciding)
Before proposing structure or a design, CONSULT the captured, enforced system model:
wiki/architecture/system-map.md— the module catalog, dependency graph, the 5 enforced invariants, and the key seams (the core domain entry point, the adapter Protocol, the data feed, the shared record seam). The[tool.importlinter]contracts inpyproject.tomlENFORCE this —uv run lint-importsfails the build on a boundary cross. Extend the existing seams; do NOT re-architect an enforced boundary. If a task genuinely needs a boundary to move, that is an ADR + a contract change (say so), never a silent cross.wiki/design-system/design-language.md(for any UI work) — the design tokens (theme.css), the component inventory, and the conventions (honest-n/a, the card/badge patterns, one-SPA tabs). Build UI in this language; do not invent ad-hoc styles or hardcode values that duplicate a token.
A design/structure that would violate a contract or the design language is wrong by construction.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

