clear-command-driven-dev
Skill9 starsChanged 29 days agoMITView on GitHub
Explains a development workflow methodology: plan with Plan Mode, copy the plan to clipboard, execute TODOs one at a time with /clear between each task, and update the plan when new information emerges. Use this skill when the user asks about "clear command driven development", "clear-driven workflow", "how to structure Claude Code sessions", or wants an overview of the plan-clear-execute cycle. For creating a new plan, see /create-plan. For updating an existing plan, see /update-plan.
In this repository2 other files it gives its agents.
- create-planSkill
- update-planSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 57
Context loading: When starting a TODO, provide only what Claude needs — the specific task (including its "Done when" and "Verify"), its …
- Distinctive: line 62
Verify before commit: Run the TODO's "Verify" step and confirm each "Done when" item holds before committing.
- Very specific: line 65
Commit cadence: Commit after each TODO completion, before /clear.
- Very specific: line 66
Update before clear: If you discover something unexpected, run /update-plan while the context is still available, then /clear after …
≈ “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
Clear Command Driven Development
A workflow methodology: persist Plan Mode output externally, execute TODOs
one at a time with /clear between each, and continuously update the plan
as new information emerges. The 1M token context window is irrelevant —
small, focused context produces better results.
Core Principles
- Plan first, code later — Always start in Plan Mode to design before implementing
- Externalize memory — Your note app is the persistent store; Claude's context is ephemeral
- One TODO, one session — Each
/clearresets context; each task gets full attention - Living plans — Plans are updated as implementation reveals new information
The Cycle
Plan Mode ──→ Copy to clipboard ──→ Paste into your notes
│
┌─────────────────────────┘
▼
Pick a TODO item
│
▼
/clear ──→ Paste TODO + minimal context ──→ Implement
│ │
│ ┌─────────────────────────┘
│ ▼
│ Discovered something new?
│ Yes ──→ /update-plan ──→ Update notes ──→ Commit ──→ /clear
│ No ──→ Commit ──→ Mark TODO [x] ──→ /clear
│ │
└──────────────┘Related Skills
| Skill | When to use |
/create-plan | Start a new plan for a feature or task |
/update-plan | Revise an existing plan based on new findings |
Execution Tips
- Context loading: When starting a TODO, provide only what Claude needs — the specific task (including its "Done when" and "Verify"), its target files, relevant references, and any shared decision table rows it cites. Do not paste the entire plan.
- CLAUDE.md synergy: Project-wide conventions belong in CLAUDE.md; task-level planning lives in your external notes. Do not duplicate.
- Verify before commit: Run the TODO's "Verify" step and confirm each "Done when" item holds before committing. Reviewers check the change against the same list.
- Commit cadence: Commit after each TODO completion, before
/clear. - Update before clear: If you discover something unexpected, run
/update-planwhile the context is still available, then/clearafter committing. - Granularity: Each TODO item should be completable in a single Claude Code session (roughly 10–30 minutes of implementation).
Anti-Patterns
- Pasting the entire plan into every session (defeats the purpose of
/clear) - Skipping the plan update phase when encountering surprises
- Mixing layers in one TODO (DDL and application code together are hard to review)
- PRs with more than one purpose (if the heading needs "and", split the PR)
- Making TODO items too large (split a big layer change by feature within that layer)
- Keeping stale plans — always reflect reality in your notes
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.

