plan-approach
Skill1 starsChanged 3 months agoMITView on GitHub
Plan an approach before any code is written. Use when the user asks how to approach something — "how should I X", "plan X", "write a plan for X", "what's the best way to Y", "design X", "think through X before coding", "give me an approach for X", "what would it take to X". Use it for any request that asks for a route rather than the work itself.
In this repository6 other files it gives its agents.
- build-newSkill
- debug-failureSkill
- explain-codeSkill
- review-diffSkill
- update-featureSkill
- write-testsSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Read the code this would touch before proposing anything.
- Very specific: line 29
If the user also asked for the work itself ("design and build X"), give the plan first and ask whether to go ahead.
- Distinctive: line 32
Prefer the approach that fits what the project already does over the one that is theoretically cleaner.
≈ “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
Plan an approach
Plan: $ARGUMENTS. If that is empty, plan what the user's last message asks for.
How to work
- Read the code this would touch before proposing anything. A plan written against what you assume is there is worth nothing. If nothing exists yet, say so and plan the layout. If the question is not about this codebase, answer it directly.
- State the goal in one sentence, and what would make it done.
- Give the steps in dependency order — each one small enough to finish and check on its own, naming the files it touches.
- Name the real decision points, with a recommendation and the tradeoff in a line each. Compare only the alternatives that are viable: usually two, sometimes one.
- Say what could go wrong: the parts most likely to break, what is already covered by tests, and what would need new ones.
Rules
- Change no files: the plan goes in your reply, unless the user asked for it as a document, and then that document is the only file you write. No implementation code. Signatures, file layout, and short sketches are fine; a finished function is not.
- If the user also asked for the work itself ("design and build X"), give the plan
first and ask whether to go ahead. Once they say yes, the rule above no longer
holds: do the work the way
/build-newor/update-featurewould. - Prefer the approach that fits what the project already does over the one that is theoretically cleaner.
- Say plainly if the request does not need a plan — some work is one obvious edit, and saying so is more useful than a five-step ceremony around it.
- Flag anything you could not resolve from the code and would need answered.
Report back
The plan is the deliverable. Aim for one screen. If the steps do not fit, go longer; never fit the screen by dropping a step, a decision or a risk. End with the first step to take.
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.

