plan-in-files
Skill1 starsChanged 41 days agoMITView on GitHub
Use for any task that spans more than a couple of steps, files, or sessions — refactors, migrations, multi-file features, anything that must survive a context reset or crash. Trigger when the user asks for a plan, says "keep track of this", "don't lose the thread", or when a task is large enough that forgetting step four is likely.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- OneWave-AI/claude-skillsword for word, first seen Sep 21, 2026
Other versions1 project uses a version of this.
Word for word · 1
- OneWave-AI/claude-skills★ 378100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
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 repository15 other files it gives its agents.
- agent-armySkill
- before-you-buildSkill
- code-review-proSkill
- cold-email-sequence-generatorSkill
- content-repurposerSkill
- expert-debate-panelSkill
- landing-page-copywriterSkill
- linkedin-post-optimizerSkill
- objection-pattern-detectorSkill
- prospect-research-compilerSkill
- sales-call-prep-assistantSkill
- scoutSkill
- seo-optimizerSkill
- session-handoffSkill
- test-coverage-improverSkill
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
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
- Distinctive: line 12
Before starting multi-step work, write the plan to disk.
- Very specific: line 14
Default location: PLAN.md at the repo root for one-off work, or docs/plans/<yyyy-mm-dd>-<topic>.md when the repo keeps plan history.
- Distinctive: line 43
Write it. Show it to the user before executing when the work is large or destructive.
- Distinctive: line 46
When reality disagrees with the plan, edit the plan.
- Distinctive: line 51
Never tick a box for work you have not verified.
- Distinctive: line 52
Never fake progress to make the plan look finished.
- Distinctive: line 53
Keep the plan under a page.
- Distinctive: line 54
If a step turns out to be five steps, split it in the file before continuing.
- Distinctive: line 55
When work finishes, write one closing line in the Log and leave the file.
- Distinctive: line 59
When picking up an existing plan file:
- Distinctive: line 62
Verify the ticked steps against reality — git log, git diff, run the tests.
- Distinctive: line 64
Start at the first genuinely unfinished step.
≈ “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
Plan in Files
A plan that lives in the conversation dies with the conversation. A plan that lives in a file survives a crash, a /clear, a laptop reboot, and a different person picking the work up on Monday.
Core Behavior
Before starting multi-step work, write the plan to disk. Update it as you go. The file is the source of truth about progress — not memory, not the chat.
Default location: PLAN.md at the repo root for one-off work, or docs/plans/<yyyy-mm-dd>-<topic>.md when the repo keeps plan history. Gitignore it or commit it — ask once, then follow the repo's convention.
Plan Shape
# <Task> — plan
**Goal:** <one sentence, in the user's words>
**Done when:** <the observable condition that ends this work>
## Constraints
- <the things that must stay true: no API break, no new deps, ships today>
## Steps
- [ ] 1. <action> — `path/to/file`
- [ ] 2. <action> — `path/to/file`
- [ ] 3. <action> — verify with `<command>`
## Notes
<discoveries, gotchas, things that turned out to be false>
## Log
- <date/time> — <what happened>Each step is one action on a named file or a named command. "Refactor the auth layer" is not a step; "move verifyToken out of api/auth.ts into lib/token.ts and update the three imports" is.
Working the Plan
- Write it. Show it to the user before executing when the work is large or destructive.
- Do one step.
- Tick the box in the file, immediately. Not at the end — a crash between step three and the end loses everything.
- When reality disagrees with the plan, edit the plan. A plan you silently abandoned is worse than no plan.
- Append surprises to Notes as you hit them. That is where the next session's time savings live.
Rules
- Never tick a box for work you have not verified. "Edited the file" is not "step done" if the step said the tests pass.
- Never fake progress to make the plan look finished. An honest half-done plan is useful; a dishonest complete one is a trap.
- Keep the plan under a page. A plan longer than the work is a symptom of avoiding the work.
- If a step turns out to be five steps, split it in the file before continuing.
- When work finishes, write one closing line in the Log and leave the file. It becomes the record of why the code looks the way it does.
Resuming
When picking up an existing plan file:
- Read it.
- Verify the ticked steps against reality —
git log,git diff, run the tests. Ticked does not always mean true. - Correct the file if it lied.
- Start at the first genuinely unfinished step.
When Not to Use This
A single-file edit, a one-line fix, a question. Writing a plan file for a two-minute change is ceremony, and ceremony is the thing this library is against. The test: if losing the session would cost more than five minutes of rework, write the file.
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.

