feature-factory
Skill0 starsChanged 3 months agoMITView on GitHub
Software-factory orchestrator that chains 7 AI specialists (researcher → story → spec → backend → frontend → test → validator) to ship one feature end-to-end through research → definition → design → implementation → verification. Run "/feature-factory <feature description>" from the root of the target project (git-managed, with a CLAUDE.md). Add "--night" for unattended overnight runs.
4 other projects use a version of this.
Adopted by
Most recently taken up by
- MShehrozJamshaid/cypher-claude-skillstheir own version, first seen Oct 20, 2026
- Enricrypto/cypher-claude-skillstheir own version, first seen Jul 12, 2026
- glebis/claude-skillstheir own version, first seen Jul 1, 2026
- JAVICHILEON/claude---Skillstheir own version, first seen Jul 1, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- glebis/claude-skills★ 410same name
- MShehrozJamshaid/cypher-claude-skills★ 3same name
- Enricrypto/cypher-claude-skills★ 2same name
- JAVICHILEON/claude---Skills★ 0same 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.
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
All artifacts are saved under factory-work/<feature-name>/ (kebab-case feature name).
- Very specific: line 44
Before bed: run Steps 4–7 back-to-back with --night, stopping at CP3 with results written to files
- Distinctive: line 48
Never stall on a question.
- Very specific: line 49
If the run stops early (rework limit, environment errors), always record how far it got and what remains in 05_final_report.md
- Very specific: line 53
After each completed run, append what worked and what hurt to references/lessons.md — the factory gets better every run.
“Common” is measured: kept by at least 6 in 10 of the 3 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
Feature Factory — Software Factory Orchestrator
Takes a rough feature description and chains 7 specialist agents (bundled with this plugin) through to a working implementation. The orchestrator (main session) sticks to design decisions, collecting approvals, and arbitrating rework — it never writes code itself.
Preconditions (if not met, stop and say so)
- The target project root has a CLAUDE.md (if not, offer to create one first)
- You are inside a git repository (so rework and rollbacks stay safe)
The chain
All artifacts are saved under factory-work/<feature-name>/ (kebab-case feature name).
| Step | Agent | Gets | Produces |
| 1 Research | codebase-researcher | feature description | current state, existing patterns, risks |
| 2 Story | story-writer | description + research | 01_story.md (acceptance criteria, edge cases, out-of-scope) |
| 3 Brief | spec-writer | approved story + research | 02_brief.md |
| 4 Backend | backend-builder | approved brief | implementation + 03_backend_summary.md (incl. API contract) |
| 5 Frontend | frontend-builder | brief + 03_backend_summary.md | implementation + 04_frontend_summary.md |
| 6 Acceptance | test-verifier | story + brief + both summaries | acceptance tests + pass/fail report |
| 7 Validation | implementation-validator | all artifacts | 05_final_report.md |
Checkpoints (get user approval before proceeding)
- ⏸ CP1 (after Step 2): approve
01_story.md - ⏸ CP2 (after Step 3): approve
02_brief.md - ⏸ CP3 (after Step 7): read the final report, decide commit/PR
Rework: acceptance criteria that fail at Step 6 and Critical/Important findings from Step 7 go back to the responsible builder for a fix + re-run. Maximum 3 rework round-trips total. Past that, stop and write the remaining issues honestly into the final report.
Iron rules
- Implementation agents (backend-builder / frontend-builder / test-verifier) run on a cheaper model (
model: sonnet, already set in each agent definition). Design, review, and integration judgment stay with the main session - The orchestrator never patches code itself. Fixes always go back to the responsible agent
- Cross-boundary bans (backend never touches UI, etc.) live in the agent definitions. The orchestrator reads each summary and verifies no boundary was crossed
- Give each agent only the artifact files it needs (never dump the whole conversation into it)
Night mode (--night)
Unattended operation, started only after CP1 and CP2 are done with a human.
- Evening: run Steps 1–3; the user approves the story and the brief
- Before bed: run Steps 4–7 back-to-back with
--night, stopping at CP3 with results written to files - Next morning: the user reads
05_final_report.mdand decides commit/PR
Additional night-mode rules:
- Never stall on a question. Choose the safe implementation and record it under "Open questions" in the final report
- If the run stops early (rework limit, environment errors), always record how far it got and what remains in
05_final_report.md - Absolutely no out-of-scope changes
Compounding
After each completed run, append what worked and what hurt to references/lessons.md — the factory gets better every run.
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.