make-todos
Skill0 starsChanged 6 months agoMITView on GitHub
Creates a structured task list for the current work. Use when the user invokes /make-todos, or when the task involves multiple steps where later steps depend on earlier ones completing first.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ask-questionsSkill
- commit-messageSkill
- review-planSkill
- smoke-testSkill
- verify-changesSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Read the project's agent instructions (e.g.
- Distinctive: line 16
Do not create a task list for:
- Distinctive: line 51
Blocked: If an item cannot proceed due to an unexpected dependency, mark it pending, note the blocker, and either create a new prerequisite …
- Distinctive: line 52
Failed: If an item's implementation does not work (e.g.
- Distinctive: line 53
Discovered work: When implementation reveals necessary work not in the original list, add new items in dependency order.
- Distinctive: line 54
Scope change: If the user changes direction, regenerate the list rather than patching a stale one.
- Distinctive: line 58
Do not create items for work that will happen automatically as a side-effect of another item.
- Distinctive: line 59
Do not front-load the entire list if scope is unclear.
- Distinctive: line 60
Do not pause after creating the list to ask for approval -- start working immediately unless the user asked to review the plan first.
- Very specific: line 64
If scope is unclear, consider /ask-questions first to clarify requirements before decomposing.
- Very specific: line 66
After all items are complete, use /commit-message to generate a commit message.
≈ “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
Make Todos
Read the project's agent instructions (e.g. AGENTS.md) before applying this skill to ensure conventions are followed.
When to apply
- User explicitly invokes
/make-todos - The work involves multiple steps with dependencies or sequencing requirements
- You are uncertain about scope and want to make the plan visible before starting
Do not create a task list for:
- Single-step tasks or trivial changes
- Conversational requests (questions, explanations)
- Work where the steps are obvious and linear (e.g. "rename this variable across the codebase")
Procedure
- Decompose the work into focused, completable units using the granularity guidance below.
- Create the task list using whatever tracking mechanism is available: a task-tracking tool (TodoWrite, TaskCreate), YAML frontmatter in plan mode, or a numbered checklist in your response.
- Start immediately -- mark the first item in-progress and begin working on it in the same response. Do not pause to announce the plan.
- Complete sequentially -- mark each item done as you finish it, then start the next. Only one item should be in-progress at a time.
Granularity
Each item should be independently verifiable -- you can tell whether it is done without checking other items.
Too coarse (bad):
- "Implement the new feature" -- covers definition, handler, registration, tests, docs
Too fine (bad):
- "Add import statement for uuid package" -- trivial, automatic side-effect
Right level (good):
- "Define tool and register it in server wiring"
- "Implement handler following existing patterns"
- "Add tests using existing test fixtures"
- "Run build/test/lint and fix failures"
- "Update project documentation"
Task naming
Name each item as an action: "Add X", "Update Y", "Fix Z", "Extract W" -- not nouns like "Error handling" or "Tests".
Mid-flight adjustments
- Blocked: If an item cannot proceed due to an unexpected dependency, mark it pending, note the blocker, and either create a new prerequisite item or ask the user.
- Failed: If an item's implementation does not work (e.g. tests fail), keep it in-progress and debug. Do not mark it done. Split into diagnostic sub-items if the problem is complex.
- Discovered work: When implementation reveals necessary work not in the original list, add new items in dependency order. Do not silently expand an existing item's scope.
- Scope change: If the user changes direction, regenerate the list rather than patching a stale one.
Anti-patterns
- Do not create items for work that will happen automatically as a side-effect of another item.
- Do not front-load the entire list if scope is unclear. Create items for what is known and add more as the work unfolds.
- Do not pause after creating the list to ask for approval -- start working immediately unless the user asked to review the plan first.
Related skills
- If scope is unclear, consider
/ask-questionsfirst to clarify requirements before decomposing. - Include a final item for
/verify-changesto run the project's build/test/lint workflow. - After all items are complete, use
/commit-messageto generate a commit message.
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.

