architect
Skill3 starsChanged 3 months agoMITView on GitHub
Staged build methodology for software projects: technical spike → design notes → build plan → stage-by-stage execution with professional-grade testing. Use when user says '/architect', 'new project', 'start a project', 'build this from scratch', wants to plan/build with structured methodology, or resumes an in-progress staged build. NOT for: quick scripts or one-off fixes.
174 other projects use a version of this.
Adopted by
Most recently taken up by
- Yeachan-Heo/oh-my-claude-sisyphustheir own version, first seen Nov 8, 2026
- caidanw/skillstheir own version, first seen Nov 8, 2026
- pascalorg/editortheir own version, first seen Nov 8, 2026
- Yeachan-Heo/oh-my-claudecodetheir own version, first seen Nov 8, 2026
- naman10parikh/pennytheir own version, first seen Nov 8, 2026
- afawfawffaf/my-skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use ecosystem-reference, fractal-delegation and 10 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.
naman10parikhSkillecosystem-reference
Quick-access directory of Claude Code ecosystem repos, skills, plugins, MCPs, and patterns. Use when improving the harness.
Also in 5 projects that use this
naman10parikhSkillfractal-delegation
Also in 5 projects that use this
tristanphvnSkillketchup-video
Catch the user up like /keel:ketchup, with a short silent video of slides for what happened and the waiting calls as text under it. Use for /keel:ketchup-video.
Also in 14 projects that use this
jaydubya818Skillfigure-it-out
Use when novel or multi-phase work needs an auditable workflow and no narrower playbook fits.
Also in 55 projects that use this
ianclearySkillprinciple-explain-the-number
Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.
Also in 29 projects that use this
eryzerzSkillmake-bot-ui
Use when building a custom UI (page, dashboard, buttons) that should wake a Grok Bot over a webhook, when the user must provide a webhook sender key, or when exposing that UI on Tailscale.
Also in 20 projects that use this
- +6 more from naman10parikh
Other versions174 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 174
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- pascalorg/editor★ 25ksame name
- backnotprop/pstack★ 3.8ksame name
- agenticnotetaking/arscontexta★ 3.5ksame name
- jsmastery-pro/skills★ 1.4ksame name
- SethGammon/Citadel★ 927same name
- moeru-ai/xsai★ 667same name
- DanMcInerney/architect-loop★ 626same name
- Hainrixz/the-architect★ 562same name
164 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 repository42 other files it gives its agents.
- AGENTS.mdAGENTS.md
- adversarySkill
- architectSkill
- eatSkill
- elevateSkill
- eyeSkill
- fortifySkill
- liquid-cat-physicsSkill
- mementoSkill
- notebookSkill
- prodevSkill
- reconSkill
- scorched-earthSkill
- vetSkill
- xraySkill
- adversarySkill
- architectSkill
- eatSkill
- elevateSkill
- eyeSkill
- fortifySkill
- liquid-cat-physicsSkill
- mementoSkill
- notebookSkill
- prodevSkill
- reconSkill
- scorched-earthSkill
- vetSkill
- xraySkill
- adversarySkill
- eatSkill
- elevateSkill
- eyeSkill
- fortifySkill
- liquid-cat-physicsSkill
- mementoSkill
- notebookSkill
- prodevSkill
- reconSkill
- scorched-earthSkill
- vetSkill
- xraySkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
30 rules. 0 are common, 18 are distinctive, and 12 are very specific to this project.
0
Common
Most of the 162 edited versions we read keep these.
18
Distinctive
This author's own choices. Worth a look.
12
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Core rule: Never work from one giant plan.
- Distinctive: line 56
If unsure: "This feels like a [tier] project — [X] major components.
- Very specific: line 58
Tier escalation mid-project: If scope expands beyond original estimate by >50%, stop before the next stage.
- Distinctive: line 70
Never skip phases. Never move forward until the current phase's gate check passes.
- Distinctive: line 76
Ask: "What's the riskiest technical assumption in this project?"
- Distinctive: line 77
If unsure, identify it yourself — state what you think the risk is and why
- Distinctive: line 99
Ask: "What am I missing? What's wrong?
- Distinctive: line 103
Do not dump 10 questions at once.
- Very specific: line 107
If _docs/recon-summary.md exists in the project root, read it before writing any design notes — it contains domain research (pitfalls, …
- Very specific: line 111
Start with 00-project-overview.md, then identify domains:
- Distinctive: line 132
Before proceeding, ALL must be true:
- Very specific: line 136
No unresolved ## Open Questions that block build planning
- Distinctive: line 139
If any check fails: Keep working on notes. Identify the specific missing item and address it before re-checking.
- Very specific: line 149
Write build_plan/_build-order.md with stage list and dependency graph
- Very specific: line 172
Every stage document (build_plan/) must include:
- Very specific: line 178
Test strategy and full test code (see references/testing.md)
- Distinctive: line 186
Read the stage document fully before writing any code
- Distinctive: line 188
Run the verification checklist
- Distinctive: line 190
All tests green, no skipped tests
- Distinctive: line 199
Plan is wrong: Stop. Update the stage document to match reality.
- Distinctive: line 205
Check if the fix changes interfaces later stages depend on.
- Distinctive: line 208
Scope creep: If adding something not in the stage doc, stop.
- Very specific: line 210
Capture lessons: When a constraint, failure, or non-obvious gotcha is discovered, append to notebook/lessons.md.
- Very specific: line 261
Read PROJECT_STATE.md — parse current phase, current_stage, last_completed_stage
- Very specific: line 262
Read notebook/_index.md and notebook/lessons.md
- Distinctive: line 264
If in execute phase, read the current stage document
- Distinctive: line 266
Ask: "Has anything changed since last session?
- Very specific: line 267
Wait for confirmation, then continue from ## Next Action
- Very specific: line 269
If PROJECT_STATE.md is missing, scan for artifacts (notebook/, build_plan/, src/) and reconstruct.
- Distinctive: line 293
Before creating any project directory or git repo, ask the user which identity context this project belongs to.
“Common” is measured: kept by at least 6 in 10 of the 162 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
Staged Build Methodology
Drive software projects through a structured lifecycle that prevents the #1 AI coding failure: losing context on a big plan and missing pieces.
When NOT to use this skill
- Task is a quick script, one-off fix, or single-file change → just build it directly
- User wants to edit existing code, not plan a new project → edit directly
- Scope is entirely clear and fits in one session without a plan → skip the methodology
- User explicitly says "just do it" without wanting a plan → respect that
Quick Reference
| Phase | What happens | Gate to pass |
| 0 | Technical spike — validate riskiest constraint | PROJECT_STATE.md with validated constraint |
| 1 | Design notes — one per domain, no code | All domains covered, user confirms |
| 1.5 | Build order — dependency graph, stage sequence | _build-order.md reviewed + approved |
| 2+Execute | Write 2-3 stages → build → repeat | All tests green, stage committed |
Full methodology: references/methodology.md · Testing obligations: references/testing.md · Stage format: references/stage-template.md
Core rule: Never work from one giant plan. Atomize into stages small enough to hold entirely in context, execute correctly, and validate before moving on.
Invocation
| Input | Action |
Bare (/architect) | Ask: "New project or resuming an existing one?" |
| With project description | Assess scope (see Scaling Tiers), then start Phase 0 |
resume or project name | Run Resume Protocol below |
Scaling Tiers
| Tier | Signal | Phases Used | Notes |
| Light | Single feature, 1-2 files, < 500 lines | Phase 0 (quick) → Phase 2 (1-2 stages) → Execute | Skip formal design notes. |
| Standard | Multi-component system, 3-8 stages | All phases | Full methodology. |
| Heavy | 8+ stages, multiple subsystems, external integrations | All phases + milestones | Group stages into milestones of 3-4. Milestone gate check between each. See references/methodology.md § Milestones. |
If unsure: "This feels like a [tier] project — [X] major components. Does that match your sense of scope?"
Tier escalation mid-project: If scope expands beyond original estimate by >50%, stop before the next stage. Re-assess tier, update _build-order.md, and confirm with the user before continuing.
Phase Sequence
Phase 0 → Technical Spike → Validate hardest constraint
Phase 1 → Design Notes → One note per domain, no code
Phase 1.5 → Build Order → Dependency graph, stage sequence
Phase 2 → Stage Documents → Write 2-3 stages ahead, not all upfront
Execute → Build Stage by Stage → Verify, commit, nextNever skip phases. Never move forward until the current phase's gate check passes.
Phase 0: Technical Spike
- Ask: "What's the riskiest technical assumption in this project?"
- If unsure, identify it yourself — state what you think the risk is and why
- Design the smallest possible test (< 50 lines) that gives a YES/NO answer
- Build and run the test
- Result:
- YES → Record the validated constraint, proceed to Phase 1
- NO → Surface the problem, discuss alternatives, re-spike if needed. Do not proceed until a constraint is validated.
Gate check: PROJECT_STATE.md exists with a validated constraint under ## Core Constraints.
If gate fails: Do not re-spike the same approach twice. If the spike produced a NO answer, the approach is blocked. Discuss alternatives with the user. If they still want to validate the same approach, ask what changed — something must have changed or the spike was invalid. Then proceed to re-spike or pivot. Do not move to Phase 1 without a validated constraint.
Phase 1: Design Notes
Mode: Planning only. Write no production code.
Conversation Protocol
For each domain:
- Draft a first-pass note based on what you know
- Present key decisions and architecture. Flag what you're uncertain about.
- Ask: "What am I missing? What's wrong? What would you change?"
- Revise based on feedback
- Confirm: "This note covers [domain]. Ready to move to [next domain]?"
Do not dump 10 questions at once. Each note is a collaborative artifact.
Pre-Build Intelligence
If _docs/recon-summary.md exists in the project root, read it before writing any design notes — it contains domain research (pitfalls, architecture decisions, user expectations) that should inform your design. Read the detail files (_docs/architecture-notes.md, _docs/pitfalls.md) as relevant per domain.
Domain Discovery
Start with 00-project-overview.md, then identify domains:
Always write:
- Data model / storage
- Core logic
- Interface (API, CLI, UI)
- Configuration
Write if applicable:
- Authentication / authorization
- External service integrations
- Error handling strategy
- Deployment / infrastructure
Write if relevant:
- Background jobs / queues, caching, real-time, search, file storage
Finish with NN-tech-stack.md and NN-project-status.md.
Gate Check
Before proceeding, ALL must be true:
notebook/_index.mdlists all notes- At least
00-project-overview.md+ one note per domain exists NN-tech-stack.mdexists with justified choices- No unresolved
## Open Questionsthat block build planning - User confirms: "Design notes look complete, ready to plan the build?"
If any check fails: Keep working on notes. Identify the specific missing item and address it before re-checking.
Phase 1.5: Spec-to-Plan Translation
- List every major component from the notes
- Map dependencies (what needs what to exist first)
- Sequence into stages: foundation first → data layer → core abstractions → implementations
- Each stage must be independently testable
- Write
build_plan/_build-order.mdwith stage list and dependency graph - Present to user for review
Gate check: _build-order.md exists with numbered stages, dependencies, and key outcomes. User has reviewed and approved.
If gate fails: Return to the stage list, identify what's missing or unclear, revise before re-presenting.
Phase 2 + Execute: Write and Build Incrementally
The Loop:
1. Write stage docs for next 2-3 stages
2. Build those stages
3. After building: does the build order still make sense?
YES → write next 2-3 stage docs, continue
NO → update _build-order.md, adjust remaining stages, continue
4. Repeat until doneWriting Stage Documents
Every stage document (build_plan/) must include:
- Overview with explicit dependencies and key outcomes
- Architecture diagram
- Directory structure (every file created or modified)
- Full implementation code (not pseudocode)
- Configuration changes
- Test strategy and full test code (see
references/testing.md) - Verification checklist (specific commands and expected outputs)
- Debugging playbook (ranked failure modes with diagnostics)
- Acceptance criteria
- Next stage preview
Executing a Stage
- Read the stage document fully before writing any code
- Build exactly what it specifies
- Run the verification checklist
- Fix anything that fails
- All tests green, no skipped tests
- Update
PROJECT_STATE.md - Commit
- Move to next stage
When Things Break
Test failure in current stage: Fix before moving on. Broken tests do not carry forward.
Plan is wrong: Stop. Update the stage document to match reality. Re-execute from the updated plan.
Prior stage bug discovered:
- Stop current stage
- Return to the broken stage and fix it
- Re-run that stage's verification checklist — all green
- Check if the fix changes interfaces later stages depend on. If yes: update affected stage docs.
- Resume current stage
Scope creep: If adding something not in the stage doc, stop. Add it to a future stage or discard it.
Capture lessons: When a constraint, failure, or non-obvious gotcha is discovered, append to notebook/lessons.md. Use /save if notebook skill is installed.
State Tracking
PROJECT_STATE.md is the single source of truth. Update after every phase transition and completed stage.
# [Project Name]
## Current Phase
phase: [0 | 1 | 1.5 | 2 | execute]
current_stage: [N or null]
last_completed_stage: [N or null]
tier: [light | standard | heavy]
## Core Constraints
| Constraint | Validated? | Result |
|------------|-----------|--------|
| [description] | yes/no | [what we found] |
## Milestones (Heavy tier only)
- [x] M1: [name] — stages 1-3 — completed [date]
- [ ] M2: [name] — stages 4-6 — in progress
- [ ] M3: [name] — stages 7-9
## Completed
- [x] Phase 0: [constraint validated]
- [x] Stage 1: [name] — [date]
- [ ] Stage 2: [name] — in progress
## Current State
**What exists:** [one line]
**What works:** [bullet list]
**What's broken:** [bullet list or "nothing"]
**Blocked on:** [description or "nothing"]
## Next Action
[Single explicit next step — not a list of possibilities]
## Build Order Reference
[Stage list with completion status]## Next Action always contains exactly one actionable step.
Resume Protocol
- Read
PROJECT_STATE.md— parse current phase, current_stage, last_completed_stage - Read
notebook/_index.mdandnotebook/lessons.md - Read
build_plan/_build-order.md - If in execute phase, read the current stage document
- Summarize: "You're in Phase [X], working on [description]. Last completed: [what]. Next action: [what]."
- Ask: "Has anything changed since last session? Any new constraints or direction changes?"
- Wait for confirmation, then continue from
## Next Action
If PROJECT_STATE.md is missing, scan for artifacts (notebook/, build_plan/, src/) and reconstruct. Ask user to confirm before proceeding.
Project Directory Structure
project/
├── PROJECT_STATE.md
├── notebook/
│ ├── _index.md
│ ├── lessons.md
│ ├── 00-project-overview.md
│ └── ...
├── build_plan/
│ ├── _build-order.md
│ ├── stage-01-foundation.md
│ └── ...
├── src/
└── tests/Identity Rules
Before creating any project directory or git repo, ask the user which identity context this project belongs to. Respect all identity separation rules from the global config.
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.

