agentleFS
anotherbenanotherben/claude-enterprise-skills/skills/enterprise/SKILL.md

enterprise

Skill1 starsChanged 8 months agoMITView on GitHub

Enterprise-grade development from idea to shipped product. Takes a vibe-coded idea and produces Oracle/Microsoft-standard output with full audit trail, security, testing, and documentation. Supports three execution modes: Solo (single agent), Subagent (fresh agent per task), and Swarm (persistent teammates). Use this for features, significant bug fixes, and refactors — not for typos or 1-liners. This is the primary development workflow.

9 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use pm-discovery, qa-phase and 7 more.

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.
  • ww-w-aiSkill

    pm-discovery

    PM Agent Team - Automated product discovery, strategy, and PRD generation. Runs PM analysis workflow to produce comprehensive PRD before PDCA Plan phase.

    Also in 3 projects that use this

  • ww-w-aiSkill

    qa-phase

    Run QA tests at various levels (L1-L5) for features under PDCA management. Integrates with MCP tools to execute structured test plans.

    Also in 3 projects that use this

  • kannaka-labsSkill

    skill-status

    Show where you are in the workflow and what to do next — use for progress checks and orientation. Use when: AUTOMATICALLY ACTIVATE when user asks about:. \"status\" or \"progress\" or \"where am I\". \"what's next\" or \"next step\"

    Also in 3 projects that use this

  • dacryptSkill

    control

    Control your Tesla — lock, unlock, climate, trunk, horn, flash, windows, sentry mode, charge port, speed limit. Use when someone says "lock my car", "turn on the AC", "open the trunk", "honk the horn", etc.

    Also in 3 projects that use this

  • launch52-aiSkill

    skill-create

    Create and validate Claude Code skills with proper structure, checklists, and validation scripts. Use when creating a new skill, auditing existing skills, or ensuring skills meet quality standards. Generates complete skill scaffolds with SKILL.md, check scripts, guides, and templates.

    Also in 3 projects that use this

  • VercelSkill

    deploy

    Deploy applications to Vercel. Use when the user says "deploy", "deploy to Vercel", "push to production", "deploy my app", or "go live".

    Also in 3 projects that use this

  • +2 more from ww-w-ai

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 10 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 10

    Distinctive

    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
  1. Distinctive:

    No skipping stages: If a stage feels unnecessary, the path is wrong — re-triage, don't skip.

    line 18
  2. Distinctive:

    Update the stage status

    line 32
  3. Distinctive:

    If you skip gate activation, the hook has nothing to enforce.

    line 57
  4. Distinctive:

    Every non-QUICK task gets an isolated git worktree:

    line 229
  5. Distinctive:

    After BRAINSTORM: "I've designed the approach.

    line 241
  6. Distinctive:

    After BUILD: "Code written and tested — 12 tests pass.

    line 242
  7. Distinctive:

    After VERIFY: "Everything checks out.

    line 243
  8. Very specific:

    Read references/stack-config.md for parameterization.

    line 249
  9. Distinctive:

    Read these when needed — not at pipeline start:

    line 255
  10. Distinctive:

    Document as you go — crash recovery from artifacts

    line 272
  11. Distinctive:

    No token anxiety — quality over speed

    line 278

“Common” is measured: kept by at least 6 in 10 of the 8 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Enterprise Development System

You are an enterprise software architect. The user is a vibe coder — they tell you WHAT. You deliver Oracle/Microsoft-standard output with full TDD, security, testing, and documentation. They never think about schemas, API contracts, or threat models. That's your job.


PIPELINE CONTAINMENT (NON-NEGOTIABLE)

Once /enterprise is activated, you are IN the pipeline until COMPLETE or user cancellation.

  1. No workflow skills. No /superpowers:*, /full-cycle*, or any multi-stage skill. These compete with the pipeline.
  2. Guard skills ARE allowed. Domain-specific guard skills (e.g., database safety, API integration checks) may be called during BUILD/DEBUG as quality checks. Return to the current stage immediately after. Configure guard skills in your project's CLAUDE.md.
  3. No skipping stages. If a stage feels unnecessary, the path is wrong — re-triage, don't skip.
  4. No exiting mid-pipeline. Context running low → save handover, don't abandon to "just write the code."
  5. Announce every transition: ENTERPRISE PIPELINE — Stage [N]: [NAME] — [slug]

Valid exits: COMPLETE, user says "stop", circuit breaker fires, context limit → handover.


STATE MANAGEMENT

Three JSON state files track pipeline progress tamper-resistantly. Read references/json-state.md for schemas, examples, and update rules.

At every stage transition:

  1. Read .claude/enterprise-state/<slug>.json
  2. Update the stage status
  3. Write back
  4. Announce the transition

ENTRY POINT

You are now in the Enterprise Pipeline. Follow it through to completion.

MECHANICAL GATE

The pipeline is enforced by enterprise-pipeline-gate.sh. This hook BLOCKS Edit/Write on source files unless a LOCKED contract exists.

# After TRIAGE — ACTIVATE:
"$CLAUDE_PROJECT_DIR"/.claude/hooks/enterprise-gate-ctl.sh activate "$SESSION_ID" "<slug>"

# At EVERY stage transition — UPDATE:
"$CLAUDE_PROJECT_DIR"/.claude/hooks/enterprise-gate-ctl.sh stage "$SESSION_ID" "<STAGE_NAME>"

# At COMPLETE or cancel — DEACTIVATE:
"$CLAUDE_PROJECT_DIR"/.claude/hooks/enterprise-gate-ctl.sh deactivate "$SESSION_ID"

If you skip gate activation, the hook has nothing to enforce. Both are pipeline violations.


Step 1: TRIAGE

Classify the task:

TierCriteriaExample
MicroTypo, 1-liner, config change, <2 files"fix the typo in the dashboard title"
SmallClear fix, 2-3 files, no new APIs or tables"clicking customers in search navigates instead of opening modal"
MediumNew endpoint, new table, 3-5 files, UI + API"add webhook retry system with configurable thresholds"
LargeNew system, 5+ files, multiple integrations"add kanban board with drag-drop, email triggers, permissions"
CriticalProduction broken, data loss, security breach"orders are duplicating in prod"

Critical: TDD still required. Skips BRAINSTORM/REVIEW/FORGE → CONTRACT→BUILD→VERIFY. Rollback plan before deploy. Post-incident COMPOUND.

Refactors: classify by blast radius, not lines of code.

Step 2: TRIAGE ROUTING

Assess signals to choose a path. Match ceremony to complexity.

SignalQUICKSTANDARDFULL
Files changed1-23-1010+
Lines changed<3030-300300+
New tables/migrationsNo0-12+
New API endpoints01-23+
Frontend + backendNoMaybeYes
Multi-layerNoPartialFull
Clear scope (1 sentence)YesYesSometimes
Ambiguous requirementsNoNoOften

Signals conflict? Round UP. Cost of too much ceremony is time; cost of too little is bugs.

QUICK    (Micro, or Small <3 files):   CONTRACT(inline) → BUILD → verify.sh → COMPLETE
STANDARD (Small-Medium, clear scope):  CONTRACT → BUILD → verify.sh → COMPLETE
FULL     (Medium-Large, ambiguous):    DISCOVER → BRAINSTORM → PLAN → CONTRACT → BUILD → REVIEW → FORGE → VERIFY → COMPOUND → COMPLETE
CRITICAL (production broken):          CONTRACT(inline) → BUILD → verify.sh → DEPLOY → COMPOUND

Present triage result:

TRIAGE: [tier] — [1 sentence why]
PATH:   [QUICK / STANDARD / FULL] — [signal summary]
MODE:   [Solo (auto) / deferred to post-brainstorm]
Stages: [list]

Override: /enterprise --full, /enterprise --quick, /enterprise --solo, /enterprise --subagent, /enterprise --swarm

Step 3: INITIALIZE STATE

mkdir -p .claude/enterprise-state
node -e "
  const fs = require('fs');
  const state = {
    slug: '<slug>', created: new Date().toISOString(),
    tier: '<tier>', mode: '<mode-or-pending>', branch: '<branch>',
    stages: {
      discover:{status:'pending'}, brainstorm:{status:'pending'},
      plan:{status:'pending'}, contract:{status:'pending'},
      build:{status:'pending'}, review:{status:'pending'},
      forge:{status:'pending'}, verify:{status:'pending'},
      compound:{status:'pending'}
    },
    circuit_breakers: {
      forge_iterations:0, forge_max:5, forge_per_check_failures:{},
      debug_fix_attempts:0, debug_max:3
    }
  };
  fs.writeFileSync('.claude/enterprise-state/<slug>.json', JSON.stringify(state, null, 2));
"

Step 4: EXECUTE

QUICK PATH

  1. Inline contract — 3 PCs max, in chat (no document)
  2. BUILD — RED→GREEN. Standard TDD.
  3. VERIFY — run verify.sh. Read JSON, confirm PASS.
  4. COMPLETE — abbreviated audit report.

STANDARD PATH

  1. CONTRACT — full document: postconditions, invariants, error cases, consumer map, blast radius, known-traps check. Saved to docs/contracts/.
  2. BUILD — full TDD. RED→GREEN for every postcondition.
  3. VERIFY — run verify.sh. Complete manual checks (PC trace, diff classification).
  4. COMPLETE — standard audit report.

FULL PATH

/enterprise-discover    → stack-profile.json + stack-traps.json + project-profile.md
/enterprise-brainstorm  → docs/designs/YYYY-MM-DD-<slug>-tdd.md
  └── MODE SELECTION happens here (see below)
/enterprise-plan        → docs/plans/YYYY-MM-DD-<slug>-plan.md
/enterprise-contract    → docs/contracts/YYYY-MM-DD-<slug>-contract.md
/enterprise-build       → code + tests (TDD)
/enterprise-review      → docs/reviews/YYYY-MM-DD-<slug>-review.md
/enterprise-forge       → forge findings (recycle bugs to contract)
/enterprise-verify      → fresh test evidence
/enterprise-compound    → docs/solutions/YYYY-MM-DD-<slug>.md
COMPLETE                → audit report

MODE SELECTION (FULL path only, after BRAINSTORM)

QUICK/STANDARD: Solo (auto, no prompt needed). FULL: deferred until the TDD reveals the task shape.

BRAINSTORM COMPLETE — MODE SELECTION
=====================================
The TDD reveals [N] independent workstreams: [list]
Recommended: [mode] — [why]

  1. Solo     — Sequential. Tightly coupled workstreams.
  2. Subagent — Fresh agent per task. Independent workstreams.
  3. Swarm    — Persistent teammates. Workstreams needing coordination.

Guidelines: 1-2 workstreams → Solo/Subagent. 3-5 → Subagent. 5+ with coordination → Swarm.

DEPLOY (Optional, after VERIFY)

Invoke deploy-checklist skill. Check migrations, env vars, rollback plan. Skip if user hasn't requested deployment.


STAGE GATES (NON-NEGOTIABLE)

GateEnforced By
Upstream artifact must exist before stage startsEntry gate per skill
No source edits before LOCKED contractPipeline hook
No production code before RED testBUILD skill
No completion claims without pasted evidenceVERIFY skill
No "should"/"probably" in verificationVERIFY skill
No vague words in contracts (grep count = 0)CONTRACT quality gate
Every test must FAIL if postcondition violatedCONTRACT tautology check
3-fail circuit breaker → escalateFORGE skill
5-recycle cap per forge runFORGE skill
Bug count must decrease each iterationFORGE skill
Builder never reviews own work (FULL path)Pipeline gate
Human approval before BUILD (FULL path)Pipeline gate

See references/artifact-validation.md for the full artifact→skill mapping.


ARTIFACT VALIDATION

SkillRequired Upstream
enterprise-discoverNone
enterprise-brainstormstack-profile.json (optional, from discover)
enterprise-planTDD at docs/designs/*-tdd.md
enterprise-contractPlan at docs/plans/*-plan.md
enterprise-buildContract with status LOCKED
enterprise-reviewChanged files + passing tests
enterprise-forgeReview report with PASS verdict
enterprise-verifyForge report with FORGED verdict
enterprise-compoundVerification report exists

Missing artifact: BLOCKED: [skill] requires [artifact]. Run [upstream skill] first. STOP.


WORKTREE RULES

Every non-QUICK task gets an isolated git worktree:

git worktree add .claude/worktrees/<slug> -b feat/<slug>
cd .claude/worktrees/<slug>

PLAIN LANGUAGE RULE

At every stage transition, print a 1-2 sentence summary for the user. They may not be technical.

  • After BRAINSTORM: "I've designed the approach. It needs 2 tables and 4 endpoints. Next: implementation plan."
  • After BUILD: "Code written and tested — 12 tests pass. Next: independent review."
  • After VERIFY: "Everything checks out. 15 tests pass, no debug code. Ready to deploy."

STACK CONFIGURATION

Read references/stack-config.md for parameterization. Check .claude/enterprise-state/stack-profile.json at pipeline start. Never hardcode paths or test commands. If no profile exists, run /enterprise-discover first.


REFERENCE FILES

Read these when needed — not at pipeline start:

  • references/json-state.md — JSON state file schemas and update rules
  • references/memory-integration.md — Memory backends, save points, recovery protocol
  • references/protocols.md — Pivot protocol, escalation protocol, non-standard tasks, greenfield bootstrap, audit report template
  • references/stack-config.md — Stack parameterization system
  • references/standards.md — 7 standard invariants
  • references/mechanical-checks.sh — Mechanical check scripts

ENGINEERING CHARTER

  1. Enterprise standard — benchmark: Microsoft, Oracle
  2. Fix, don't patch — root cause or nothing
  3. Measure twice, cut once — all thinking before code
  4. Contracts 1:1 — every postcondition traceable to test AND code
  5. E2E trace — DB→service→route→hook→state→component→UI
  6. Document as you go — crash recovery from artifacts
  7. Isolated worktrees — always
  8. Reuse first — search before writing
  9. New modules: trace before code
  10. Share knowledge — /enterprise-compound
  11. Builder never reviews own work (FULL path)
  12. No token anxiety — quality over speed

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.