naga-factory
Skill1 starsChanged 2 months agoMITView on GitHub
Naga Codex factory workflow — 4-beat isolate/build/prove/ship pipeline orchestrating new-feature, code-structure, evidence-driven-testing, before-and-after, greploop/greploop-apps, and unslop. Use when starting any new feature, fixing a bug, or shipping a PR; when the user says "factory", "naga workflow", "start new task", or wants the full Isolate-Build-Prove-Ship cycle with before/after proof and 5/5 Greptile.
- Commits and pushes
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- naga-factory/AGENTS.mdAGENTS.md
- before-and-afterSkill
- code-structureSkill
- evidence-driven-testingSkill
- greploop-appsSkill
- greploopSkill
- new-featureSkill
- unslopSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 12
Isolate —
/new-feature: Every task starts in a fresh Git worktree branched from origin/main. - Very specific: line 16
Prove —
/evidence-driven-testing: Verify with repo checks plus runtime evidence. - Distinctive: line 24
Never commit directly to main.
- Very specific: line 25
Never reuse another agent's worktree/branch/uncommitted work.
- Very specific: line 26
Only --force-with-lease on your own branch.
- Distinctive: line 32
Keep changes scoped to the assigned task.
- Very specific: line 35
Commit with clear message, rebase onto origin/main, rerun checks.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Naga Codex Factory
Master skill that ties the 7-piece factory pipeline into the 4-beat workflow from AGENTS.md. Each beat delegates to a focused sub-skill.
Workflow — 4 beats
- Isolate —
/new-featureEvery task starts in a fresh Git worktree branched fromorigin/main. Never build onmain. Scope-check open PRs (gh pr list+gh pr diff <n> --name-only) before creating the branch. - Build —
/code-structureEnforce two-layer separation: actions/boundaries orchestrate the "why/when" (business rules, auth, state transitions), service layer owns the reusable "how" (provider SDKs, command execution, health checks) with explicit inputs and structured returns. - Prove —
/evidence-driven-testingVerify with repo checks plus runtime evidence. Record withscripts/evidence.py(or headlessbefore-and-after/ Playwright). Capture the before failure before fixing, then the after success. Outputevidence.mp4,report.md,manifest.json. - Ship —
/before-and-afterthen/greploopEmbed before/after proof in the PR description (before-and-after <before> <after> --markdown+ upload adapter). Run/greploop(or/greploop-appsfor huge PRs over Greptile file limit) until 5/5 with zero unresolved comments. Present PR URL.
Apply /unslop to every human-facing text (commits, PR title/body, docs, comments, replies) before posting.
Multi-agent rules
- Never commit directly to
main; one worktree + one branch per task/agent. - Never reuse another agent's worktree/branch/uncommitted work.
- Only
--force-with-leaseon your own branch; never force-push main. - Regenerate lockfiles, don't hand-merge.
- Worktrees don't isolate ports/DBs — verify
lsof -i :<port>serves your process.
Completing a task
- Keep changes scoped to the assigned task.
- Run repo checks.
- Assemble before/after pairs from evidence.
- Commit with clear message, rebase onto
origin/main, rerun checks. - Push (
git push -u origin <branch>;--force-with-leaseafter rebase). - Open PR with what/how-tested/before-after/risks, run title/body through
/unslop. - Run
/greploopuntil 5/5. - Present PR URL. Do not merge unless explicitly instructed; keep worktree until merged/closed.
Skill sources
| Skill | Source |
new-feature, code-structure, evidence-driven-testing | this factory |
before-and-after | vendored vercel-labs/before-and-after (@vercel/before-and-after) |
greploop | vendored greptileai/skills |
greploop-apps | local variant of greploop for huge PRs |
unslop | vendored cursor/plugins (pstack) — frontmatter edited for auto-invocation |
Sub-skills bundled
Invoke individually or let this orchestrator delegate:
/new-feature— isolated worktree fromorigin/main/code-structure— service-layer architecture/evidence-driven-testing— annotated screen recording + headless path/before-and-after— visual diff + markdown upload/greploop— loop until Greptile 5/5/greploop-apps— same but with@greptile-appsfor oversized PRs/unslop— strip AI tells from human-facing prose
See AGENTS.md for the canonical workflow text and README.md for per-skill details. Drop AGENTS.md into any repo to enforce the workflow.
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.