atlas-fix
Skill0 starsChanged 10 days agoMITView on GitHub
Gets a failing build, type check or linter back to green with the smallest changes, one error at a time, re-running after each. Stops and says why instead of guessing: same error twice, more errors than before, a missing dependency, a change of design. Use for "atlas fix", "fix the build", "it does not compile", "type errors", "lint is red", "CI is failing" — in any language.
In this repository10 other files it gives its agents.
- atlas-artifactSkill
- atlas-commitSkill
- atlas-helpSkill
- atlas-memorySkill
- atlas-organizeSkill
- atlas-recapSkill
- atlas-reviewSkill
- atlas-searchSkill
- atlas-silentSkill
- atlasSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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 43
After the first run: more than five errors, or a build that takes longer than about twenty seconds, or a CI log to work from: delegate to …
≈ “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
One error, one smallest change, run again. Never a rewrite to make a compiler quiet.
What to read
Everything below lives in the fix/ folder beside this file.
fix/method.md, always: how to find the command, the order to fix in, when to stop, what the report looks like.- One file for the toolchain that is failing, chosen by the project's files:
typescript.md(package.json,tsconfig.json),python.md(pyproject.toml,requirements.txt,manage.py),go.md(go.mod),rust.md(Cargo.toml),java-kotlin.md(pom.xml,build.gradle,build.gradle.kts),swift.md(Package.swift,*.xcodeproj),cpp.md(CMakeLists.txt,Makefile),csharp.md(*.csproj,*.sln),dart.md(pubspec.yaml). No file for the toolchain: work frommethod.mdalone and say so in the report.
What is being fixed, by the words of the request
"build", "compile", nothing named → the build command.
"types", "typecheck" → the type checker only.
"lint" → the linter only; formatting errors are fixed by the project's formatter command, not
by hand, run on the files this session changed and no others. The linter's own rules and
configuration are never edited to pass.
"CI" → the first failing step of the workflow file, run locally; a step that cannot run
locally is named and left.
"tests" → only what stops the tests from running: import errors, compile errors, fixtures that
do not load. A test that runs and fails on an assertion is a behaviour, not a build error: one
line each in the report, handed back. Asked outright to make failing tests pass: method.md,
last section.
Where the work goes
After the first run: more than five errors, or a build that takes longer than about twenty
seconds, or a CI log to work from: delegate to the atlas-fixer subagent, when it exists,
with the project root, the exact command, and the absolute paths of fix/method.md and the
toolchain file (they live beside this file). It runs the whole loop where the logs stay and
hands back the report and the list of files it changed; print the report unchanged and say
the files are already edited. One or two errors you can see: fix here.
Who runs the command
The atlas-runner subagent, when it exists: give it the exact command, take back the verdict
and the error lines, never the log. Without it: run the command with output to a file in the
scratch directory and read only the error lines from that file. Either way the log does not
enter this conversation.
Form
The active compression level governs the prose. It never overrides the report format in
method.md: the shape stays, the words inside it shorten. Error strings are quoted exactly.
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.

