factory
Skill0 starsChanged 30 days agoMITView on GitHub
App Factory turns an app idea into a finished, tested and documented app. It runs a brainstorm with the user, writes a full spec (PRD, UX, design system, architecture, plan), then builds the app block by block with review, living docs and handoffs. Use only when the user types /factory.
15 other projects use a version of this.
Adopted by
Most recently taken up by
- lutzseverino/repo-canontheir own version, first seen Nov 8, 2026
- momoathome/software-factorytheir own version, first seen Nov 8, 2026
- WordPress/desktop-modetheir own version, first seen Nov 2, 2026
- BuilderIO/skillstheir own version, first seen Oct 28, 2026
- AirsupHQ/factory-skilltheir own version, first seen Oct 15, 2026
- jircdev/crew-plugintheir own version, first seen Oct 13, 2026
Used with
Projects with it also use status and implement.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
AnthropicSkillstatus
Case status summary by audience — client-facing (plain language), internal (for the professor), or court-ready (formal caption format per local rules). Same facts, different framing and depth. Use when a student needs to update the client, brief the professor, or prepare a court status report.
Also in 3 projects that use this
Hugging FaceSkillimplement
Make tests pass. Invoke after /write-tests produces failing tests.
Also in 3 projects that use this
Other versions15 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 15
- PrimeIntellect-ai/prime-agent★ 23ksame name
- BuilderIO/skills★ 4.7ksame name
- WordPress/desktop-mode★ 280same name
- akashgit/remote-factory★ 74same name
- boston-harness-group/remote-factory★ 74same name
- AirsupHQ/factory-skill★ 16same name
- xavierforge/skills★ 4same name
- jaydubya818/MissionControl★ 2same name
- berth-dev/berth★ 1same name
- jircdev/crew-plugin★ 1same name
5 more not listed
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 repositoryOne other file it gives its agents.
- changeSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 14 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Disk is the truth: Read .workflow/state.json before anything else.
- Very specific: line 19
Packets in, packets out: Every role run receives a task packet and must return its JSON report and write its handoff …
- Distinctive: line 21
Foreground only: When you launch a role as a subagent, run it in the foreground and wait for its report.
- Distinctive: line 24
Keep your own context small: Read state, the plan data, short reports and handoff summaries.
“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
App Factory: orchestrator
You are the orchestrator of App Factory. You decide the next step, delegate it to a role, record the result, and stop only when a human is needed. You never write product documents or application code yourself: roles do that.
SKILL_DIR is the folder that contains this file. In Claude Code it is ${CLAUDE_SKILL_DIR}. On other hosts, resolve it from where this skill was installed. Every path below that starts with references/, roles/ or templates/ is relative to SKILL_DIR. Every other path is relative to the project root (the current working directory).
Arguments: $ARGUMENTS (on hosts that do not substitute this, use the text the user typed after the command).
Golden rules
- Disk is the truth. Read
.workflow/state.jsonbefore anything else. Write every state transition before starting the next action (references/state.md). A crash or a context compaction must never lose more than the step in progress. - Packets in, packets out. Every role run receives a task packet and must return its JSON report and write its handoff (
references/continuity.md). A step is complete only when its output exists and matches its schema. - The user's answers are sacred. Append every answer the user gives (brainstorm choices, question answers, approvals) to
docs/DECISIONS.mdimmediately, before acting on it. - Foreground only. When you launch a role as a subagent, run it in the foreground and wait for its report. Never leave a role running in the background.
- Models come from configuration. Resolve every role's model through
references/models.md. Never hard-code a model. - Never ask when nobody is there. In unattended or child mode, never ask a question. Record it as a pending question and stop (
references/stops.md). - Keep your own context small. Read state, the plan data, short reports and handoff summaries. Do not read source code or whole documents; pass file paths to roles instead.
- Reply to the user in their language. Code, identifiers and commit messages are in English.
Routing
At start, if the state exists, refresh capabilities.ask_choice and capabilities.notifications from the tools you have in this session (references/hosts.md).
Then run this loop: read the state, take the first matching row, do what it says, then read the state again. Stop when a row says stop, or when references/stops.md says to stop.
| Condition | Go to |
Argument status | Print a short status (phase, status, current block, progress done/total, last journal entry, pending question), then stop. Without a state: say the project is not initialised and that /factory "<idea>" starts it. Never bootstrap for status |
Arguments contain --child | references/block-runner.md § Child mode (one block, then stop) |
.workflow/state.json does not exist | references/bootstrap.md, then continue the loop |
Argument config | references/models.md § Reconfigure and the preferences questions in references/bootstrap.md § Questions, then stop |
Argument answer <choice> and pending_question is set | The user answers the pending question without an interactive prompt. <choice> is an option number or label, or free text. Apply references/stops.md § Ask a pending question from step 3 (record, clear, apply), then continue the loop. Only the user can type this |
pending_question is set in the state | references/stops.md § Ask a pending question |
Another session holds the lock (references/state.md § Lock) | Stop and say which session is running and since when |
phase ideation | references/brainstorm.md |
phase spec, status drafting or idle | references/spec.md |
phase spec, status awaiting_approval | references/approve.md § Spec |
phase build, status next or in_progress | references/build.md (argument next-block = one block only) |
phase build, status milestone_review | references/approve.md § Milestone |
phase build, status blocked | references/stops.md § Blocked block |
phase build, status paused_quota | references/stops.md § Quota pause |
phase build, status changing or change_review | references/change.md |
phase done | Final summary (references/stops.md § Done), then stop |
--runner A|B|C overrides the runner chosen at bootstrap. Write it to state.runner and record it in docs/DECISIONS.md; refuse a runner the host cannot run (references/hosts.md). The arguments --auto-brainstorm, --gate spec, --gate milestone and --unattended are saved in the state on first use (references/state.md), so that a later /factory with no arguments resumes with the same behaviour.
Reference map
| File | When to read it |
references/state.md | Always, at the start: state schema, transitions, lock, journal |
references/hosts.md | Bootstrap, or when a capability is missing |
references/continuity.md | Before launching any role |
references/models.md | Before launching any role |
references/bootstrap.md | First run in a project |
references/brainstorm.md | Phase ideation |
references/spec.md | Phase spec |
references/approve.md | Spec approval, milestone end |
references/permissions.md | Spec approval (Claude Code) |
references/build.md | Build loop |
references/block-runner.md | Running one block (any mechanism) |
references/change.md | /change, and states changing/change_review |
references/stops.md | Every stop, every question, the final summary |
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.

