start-project
Skill1 starsChanged 15 days agoMITView on GitHub
Start a new project from an empty or near-empty directory using the stack the user named. Use when there is no existing app, or the user says new project, scaffold, greenfield, or from scratch. Do not invent a stack they did not choose. Do not run this inside a repo that already has a product.
9 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- rpraharaj/ai-sdlcedited, first seen Nov 1, 2026
- jasonku09/altitude-skillstheir own version, first seen Nov 1, 2026
- 26zl/universal-agent-skillstheir own version, first seen Oct 13, 2026
- jcarlosrodicio/normatheir own version, first seen Oct 9, 2026
- NirDiamant/Agentic_Engineeringtheir own version, first seen Sep 21, 2026
- joelaharris98/legal-ai-engineering-labtheir own version, first seen Sep 3, 2026
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- rpraharaj/ai-sdlc★ 040% same
Wrote their own version · 8
- NirDiamant/Agentic_Engineering★ 34same name
- jasonku09/altitude-skills★ 27same name
- 26zl/universal-agent-skills★ 2same name
- joelaharris98/legal-ai-engineering-lab★ 2same name
- jcarlosrodicio/norma★ 1same name
- markyafi1991/project-delivery-pack★ 1same name
- pi247/classic-PM-skills★ 1same name
- yuuuguuu123-lgtm/start-project-skill★ 1same 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.
In this repository34 other files it gives its agents.
- AGENTS.mdAGENTS.md
- templates/AGENTS.mdAGENTS.md
- commitSkill
- cut-releaseSkill
- data-changeSkill
- debugSkill
- engineering-barSkill
- explain-codeSkill
- explore-codebaseSkill
- fix-ciSkill
- handoffSkill
- implement-changeSkill
- incident-triageSkill
- lean-reviewSkill
- open-prSkill
- performanceSkill
- production-passSkill
- ready-to-mergeSkill
- record-decisionSkill
- refactor-safelySkill
- resolve-conflictsSkill
- revert-changeSkill
- review-changeSkill
- security-passSkill
- shape-workSkill
- ship-notesSkill
- upgrade-depsSkill
- verify-uiSkill
- write-architectureSkill
- write-designSkill
- write-dev-tasksSkill
- write-docsSkill
- write-prdSkill
- write-testsSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 5 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
3
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
- Distinctive: line 12
Create the smallest runnable project the user asked for.
- Very specific: line 16
Stop and use explore-codebase / implement-change if the workspace already has an application: a source tree with a manifest, lockfile, or …
- Distinctive: line 18
If they asked to add an app inside a monorepo, stay in the package they named and match sibling packages.
- Distinctive: line 22
Use the language, kind (app / lib / CLI / service), and tools the user named.
- Very specific: line 24
If any of those are missing, ask one blocking question that lists the gaps.
- Very specific: line 26
Prefer the official init for the named toolchain (cargo init, go mod init, dotnet new, language-native module init).
- Distinctive: line 38
Do not commit secrets. Do not commit unless asked.
- Distinctive: line 42
Run Install if needed, then Test, then Run as a one-shot if that is cheap.
“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.
Files
Start project
Create the smallest runnable project the user asked for. Do not pick a framework for them. Do not add a product feature beyond hello + the checks needed to prove it runs.
Is this the right skill?
Stop and use explore-codebase / implement-change if the workspace already has an application: a source tree with a manifest, lockfile, or more than bootstrap files (README, LICENSE, .gitignore, AGENTS.md).
If they asked to add an app inside a monorepo, stay in the package they named and match sibling packages.
Stack
Use the language, kind (app / lib / CLI / service), and tools the user named.
If any of those are missing, ask one blocking question that lists the gaps. Do not recommend a trendy stack. Do not run a third-party create-* scaffolder unless they named that exact tool.
Prefer the official init for the named toolchain (cargo init, go mod init, dotnet new, language-native module init). Match the package manager they named.
Create
- Init the project with that official tool (or a minimal manifest if no init exists).
- One entry point that runs.
- One test that fails if the entry is broken, in the runner that toolchain already uses.
- A root
READMEwith Install / Run / Test from commands you actually used. Preserve existing content; update only relevant sections. Do not run an initializer that overwrites existing files without first arranging a safe merge. - If
AGENTS.mdexists, preserve its prose and update only the managed sections usingexplore-codebaseStep 6. If absent, copy../explore-codebase/references/agents-template.md, else../../templates/AGENTS.mdwhen available. If neither is readable, create only the repository facts and glossary, and report the missing defaults template. Fill facts from commands you ran. Never overwrite an existing instructions file with a starter template.
Load engineering-bar (paths in that skill). Climb Minimum path. No extra dependencies, no extra folders, no CI you were not asked for.
Do not commit secrets. Do not commit unless asked.
Prove
Run Install if needed, then Test, then Run as a one-shot if that is cheap. Keep the output. Unverified if you could not run them.
Done
- Stack and kind
- Files created
- Commands run and results
- Path of
AGENTS.md/README
Next is write-prd if the product is still open, shape-work or implement-change for the first real feature — not a drive-by architecture.
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.