agentleFS
javierbreajavierbrea/eslint-plugin-boundaries/.agents/skills/implement-change/SKILL.md

implement-change

Skill1k starsChanged 2 days agoMITView on GitHub

Runs a change of the code from its goal to its review — the main session agrees it with the user and coordinates, a coder subagent implements each task, and other subagents review it before it is shown. Use for every change of the code, the tests or the configs of the repository, whoever asks for it and however small

18 other projects use a version of this.

Trending

More talk lately.about the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 0 mentionsApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 3 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 11 mentionsNov 2026: 36 mentions
Dec 2025Jun 2026Nov 2026
Mentions a month of the javierbrea/eslint-plugin-boundaries repository that our weekly crawl found. There is no adoption data for this file yet.

Most recently taken up by

Who's talking about javierbrea/eslint-plugin-boundaries

The biggest 5 of 52 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 52

Talked about on GitHub and Bluesky (52 times).

On GitHub · 50

45 more on GitHub

Used with

Projects with it also use audit-test-suite, change-data-safely 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.
  • yaohuangguanSkill

    audit-test-suite

    Find where an existing test suite is weak and decide what to test next, ranked by risk. Use this when asked to add missing tests, find gaps, raise coverage, assess test quality or see which areas are thin, even if the request only mentions a coverage percentage.

    Also in 3 projects that use this

  • yaohuangguanSkill

    change-data-safely

    Change a database schema or stored data without breaking running code or losing data. Use this for any migration, new or changed column, backfill, bulk update, delete, index, or change to a stored format, including in small projects. Data mistakes are often impossible to undo.

    Also in 3 projects that use this

  • BanKnightSkill

    plan-change

    将已完成 spec/design 的 change 拆成可执行 plan 与 tasks。用户要把设计落成执行计划、任务清单或实现步骤时使用。

    Also in 6 projects that use this

  • SauravSrivastavSkill

    debug-failure

    Diagnose a failing test, command, or runtime behavior using evidence and produce a minimal verified fix.

    Also in 3 projects that use this

  • MicrosoftSkill

    fix-flaky-test

    Reproduces and fixes flaky or quarantined tests. Tries local reproduction first (fast), then falls back to CI reproduce workflow (reproduce-flaky-tests.yml). Use this when asked to investigate, reproduce, debug, or fix a flaky test, a quarantined test, or an intermittently failing test.

    Also in 3 projects that use this

  • srclightSkill

    refactor-safely

    This skill should be used when the user asks to "refactor", "rename", "move", "change a function signature", "extract a method", "delete unused code", "clean up", or any modification that could break callers or dependents. Provides a pre-change safety checklist using srclight's graph tools to assess blast radius before touching code.

    Also in 5 projects that use this

  • +6 more from yaohuangguan

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 17 edited versions we read keep these.

  • 1

    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:

    Understand the change: Read the request, the phase of the plan when the session belongs to one, and the AGENTS.md of each project the …

    line 32
  2. Very specific:

    Stop after the third round that still leaves a blocking finding, and tell the user what is left and why it does not converge.

    line 45

“Common” is measured: kept by at least 6 in 10 of the 17 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

Implement a change

When to use

For every change of a file under packages/, examples/, support/, test/, scripts/, .github/workflows/ or .husky/, or of a config at the root. A release is run by the release skill instead.

Sizing the change

The changeSizeDesigned byAgreed with the userReviewed by
Stays inside one element of one project, after a pattern the project already hasContainedThe main sessionOnly what the request leaves openreviewer
Crosses a boundary between projects or between the elements of one, after an existing patternArchitecturalarchitectAlways, as references/agreeing-the-change.md saysreviewer and architecture-reviewer
Adds a package, an element type, a public export, an entry point, a dependency between packages or a pattern, or changes an invariant or a decision of the planArchitectural and newarchitect, then skepticAlways, the same wayreviewer and architecture-reviewer
  • When in doubt, the change is of the larger size.
  • A contained change that turns out to cross a boundary stops, and starts again as architectural.

Who does what

  • The main session coordinates the whole change and keeps its context, which no subagent sees. It delegates the code, never the judgment of the design, and does not write the code itself.
  • A subagent sees only its brief. Whatever it needs from the conversation goes in it.
  • A coder never commits and never decides a placement. The other subagents edit nothing.
  • Only the main session commits, what passed its review, and only when the user or the plan asks for it. Before each commit, git diff --cached --name-only lists only the files intended.

Procedure

  1. Understand the change. Read the request, the phase of the plan when the session belongs to one, and the AGENTS.md of each project the change touches. Size it.
  2. Agree it. An architectural change follows references/agreeing-the-change.md: no code is written before the user approves its goal, its design and how it is accepted. A contained one is placed by the session after the AGENTS.md of its project, and the user is asked only what the request leaves open.
    • Either way, this step ends with the goal, how the change is accepted, and a list of tasks, each with its files, the interfaces it crosses and its tests. The CHANGELOG.md entries that .agents/rules/changelog-and-versioning.md requires belong to the tasks.
  3. Implement it. Delegate each task to a coder, with the brief below. Two tasks run at once only when they share no file and neither of their projects depends on the other; the others run in the order of the design.
  4. Read what each coder returns. A deviation from the brief is accepted with its reason or sent back. A question is answered from the docs or taken to the user, never guessed.
  5. Run the checks: pnpm nx affected -t check:all --base=<the branch the change started from>; pnpm lint and pnpm check:spell at the root, which no Nx target runs; and, for each changed source file of a package with test:mutation, specs excluded, the run scoped to it that .agents/rules/mutation-testing.md gives. A failing check goes back to a coder, and so does each surviving mutant, until it is killed or commented as that rule says.
  6. Review the whole change yourself, reading its full diff against the design:
    • Each file is where the design placed it, and the change crosses only the interfaces of the design.
    • Nothing changed outside the tasks, such as a script or a broken behavior left behind by a coder.
    • The agent docs and the README.md of each project touched are still true, or are updated now.
  7. Have it reviewed by the subagents its size asks for, at once, each with what the change is for, how it is accepted, its design and the range of the diff. Never give them what the coder concluded or what you expect them to find.
  8. Check each finding before acting on it, yours included, as "Checking a finding" says.
  9. Fix what is true. Send the true findings to a coder, each with its evidence, run the checks of step 5 again, and have the reviewer that raised each one review only what the fix changed.
    • Stop after the third round that still leaves a blocking finding, and tell the user what is left and why it does not converge.
  10. Check each item of how it is accepted.
  11. Report what "The report" lists.

The brief of a coder

  • The goal, and what it is for, in two sentences.
  • The files to create or edit, each with its project and its element, and the example to copy as a path.
  • The interfaces the task crosses, with their exact shape.
  • What it must not touch, such as the files of a task that runs at once.
  • The tests: each behavior to cover, and each existing test to rewrite or move.
  • A folder of its own for scripts and temporary files, outside the repository, so that coders running at once never share one.
  • How it is accepted: the commands that must pass.
  • In a round of fixes: each finding with its file:line and its evidence.

Checking a finding

A finding is a claim about the code, and whoever raises it can be wrong. Check it by the cheapest way that can show it false:

The finding saysCheck it by
The code does something wrongA test or a command that shows the wrong result
A rule or an invariant is brokenReading the rule, and the line that breaks it
A boundary is crossedThe lint of the project, or its eslint.config.* when the lint cannot see it
A test proves nothingBreaking the behavior and running the test
A case is not handledRunning that case
A library or a tool behaves in some wayIts source, its types or a run of it, never what it is known for

Each finding ends as one of three:

  • True: it goes to the round of fixes, or changes the design.
  • False: it is dropped, and kept with its evidence for the report.
  • A matter of taste, or a change of what was agreed: decided from the docs when they settle it, and asked to the user when they do not.
  • A check that changes a file is undone, such as a behavior broken to try a test.
  • Never drop a finding because fixing it is expensive, and never fix one that was not checked.

The report

  • What changed, by project, and the design decisions behind it.
  • Each item of how it is accepted, with its result.
  • The findings fixed, and those dropped, each with its evidence.
  • The checks run, with their real result.
  • What is left: what could not be checked, what was found outside the change, and what waits for the user.

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.