project-contract
Skill0 starsChanged 3 months agoMITView on GitHub
Create, validate, and maintain a compact opt-in technical contract for a project's accepted decisions and invariants. Load when setting up a project contract, or when a project AGENTS.md declares an active or draft Project Contract.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- data-workSkill
- frontend-designSkill
- geospatialSkill
- git-workflowSkill
- python-devops-stackSkill
- research-protocolSkill
- review-protocolSkill
- setupSkill
- test-qualitySkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
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
- Very specific: line 53
Establish: run the short interview (purpose, key risks, accepted decisions, required evidence), write a draft, present for approval, then …
- Distinctive: line 72
No project-scope prose, source-code explanations, or dependency-version lists.
≈ “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
Project Contract
A compact, opt-in, Git-versioned record of a project's accepted technical decisions and invariants. It is local, normative once active, and never activated automatically.
When to load
- The user asks to set up, validate, or amend a project contract ("set up a project contract").
- A project-local
AGENTS.mddeclares a Project Contract marker withStatus: activeordraft. - A plan or review needs to verify contract compliance.
Compact layout
AGENTS.md ## Project Contract
├─ Status: active | draft | opted-out | not-required
├─ Manifest: <path to TECHNICAL_CONTRACT.md>
├─ Activation reason: <manual | short reason>
└─ Opt-out reason: <required when Status=opted-out>
TECHNICAL_CONTRACT.md # single manifest, always read when activeThe manifest records status, authority, accepted decisions, risk controls with required evidence, and open items. There is no module tree and no docs/decisions/ requirement.
Activation rules
| Signal | Status |
| No request, quick/standard project | not-required |
| Explicit "set up a project contract" | draft → user approval → active |
| User declines after offer | opted-out with a reason |
A contract is never silently active: it becomes active only after user approval of an explicit draft with no blocking open items.
Structural validation
python3 <skill>/scripts/check_project_contract.py --root <project> [--json]Exit codes:
0— validnot-required, validactive, or explicitopted-outwith reason.2— missing required active artifacts,draft, marker/manifest mismatch, or unsafe manifest path.
The checker proves shape and references only, never semantic validity.
Lifecycle
- Establish: run the short interview (purpose, key risks, accepted decisions, required evidence), write a
draft, present for approval, then setactive. No external write-back. - Amend: a material contract change is a plan-named, user-approved change; code follows in the same approved package.
- Consume: plans and reviews read the relevant sections and do not rewrite the contract.
- Retire: downgrading or completing a project never deletes contract history.
Conflict protocol
Contract, code, config, and the brief disagreeing is a conflict, not a silent-winner decision. Do not rewrite the contract to match code, and do not ignore the contract for new work. Return to plan with:
STOP — Project Contract Gate
Observed: <missing artifact or exact contradiction>
Boundary: <contract section and plan task>
Decision needed: <smallest planning decision>Content rules
- Record only decisions, invariants, risk controls, and required evidence.
- No project-scope prose, source-code explanations, or dependency-version lists.
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.

