quality-bar
Skill1 starsChanged 4 days agoMITView on GitHub
Write a project's quality bar down once (CONSTRAINTS.md) - the checks that must pass, with real commands and numbers - so agents can't quietly lower it. Use when starting a project or big feature, before letting agents run unattended, when CI checks are unclear, or when the user wants agents to stop shipping junk.
In this repository20 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agents-mdSkill
- app-iconSkill
- buildSkill
- cloneSkill
- debugSkill
- design-systemSkill
- docker-optimizeSkill
- improve-loopSkill
- mac-appSkill
- mac-designSkill
- product-briefSkill
- refactorSkill
- researchSkill
- reviewSkill
- start-hereSkill
- stop-overthinkingSkill
- ui-directorSkill
- user-storiesSkill
- ux-designSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 21
When a check fails mid-task, block or warn?
- Distinctive: line 61
Every enforced row names the command that gives the verdict.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Quality bar
The outcome: one short file that says what "good enough to ship" means for this project, with a command for every rule. An agent writes more in an afternoon than anyone reads in a week, so the bar has to live in the repo, not in someone's head.
1. Read before you ask
Find the stack, the test runner, linters, type checker, current coverage, CI, and existing agent instructions. Report what you found in two lines, then ask only what is left.
2. Ask four questions, each with a default
- Beyond the floor, what should be enforced: test coverage on new code, security scanning, performance budgets, accessibility, architecture boundaries? Default: coverage and security.
- When a check fails mid-task, block or warn? Default: block on the floor, warn on the rest for the first two weeks.
- Target numbers, or measure today and hold the line? Default: measure and hold.
- The slowest check you'll wait for at the end of a task? Default: about 90 seconds; slower checks run in CI.
In an unattended run, don't ask: apply the floor and the defaults and say so in the report.
3. Write CONSTRAINTS.md
# Constraints
## Floor (always)
- No new suppression comments (ts-ignore, eslint-disable, noqa, swiftlint:disable) without a reason.
- No unimplemented stubs, empty catch blocks, or skipped tests without a reason in the commit.
- No secrets in source. No duplicate of an existing helper, component or token.
- Tests come before the code they prove, and fail first; a test is never edited to match a bug.
- Every "done" cites its evidence: the command and its output, `file:line`, or a screenshot.
- This file is never weakened to make a change pass.
## Enforced
| Dimension | Rule | Checked by | Runs at |
|---|---|---|---|
| Types | zero errors | `<command>` | every change |
| Lint | zero errors | `<command>` | every change |
| Tests | all pass; changed logic is covered | `<command>` | task end, CI |
| Security | nothing high or above | `<command>` | CI |
| Accessibility | zero serious issues | `<command or manual check>` | before release |
| Performance | <budget, e.g. first frame under 100 ms> | `<command>` | before release |
## Measured, must not get worse
| Metric | Today | Direction |
|---|---|---|
## Exceptions
| Rule | Path | Reason | Expires |
|---|---|---|---|Every enforced row names the command that gives the verdict. A number with no command is a wish. Raise a measured number only in its good direction (a ratchet).
4. Point every agent at it
One line in AGENTS.md: "Read CONSTRAINTS.md before writing code. Run its checks before calling
a task done. Never weaken it to make a change pass."
Done when
CONSTRAINTS.md exists, every enforced row's command runs and passes today (or its current
number is recorded), and AGENTS.md points to it.
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.

