give-plan
Skill119 starsChanged 6 months agoMITView on GitHub
Use when the user wants a written, reviewable plan or spec produced before coding starts. Triggers on: mapping out changes without implementing, thinking through risks of upgrades or migrations, evaluating approaches before committing to one, writing specs for team review, phasing work into stages, or any request that explicitly defers coding ('don't implement yet', 'before we build'). The distinguishing signal is that the user wants a plan artifact — not implementation, not a conversational answer. MUST activates inside Claude's native plan mode to have a better planning behavior.
In this repository22 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- agent-browserSkill
- askSkill
- cookSkill
- create-docSkill
- diagnoseSkill
- discussSkill
- docs-seekerSkill
- fixSkill
- frontend-designSkill
- media-processorSkill
- optimus-primeSkill
- prime-syncSkill
- review-codeSkill
- self-evolveSkill
- skill-creatorSkill
- testSkill
- backend-fastapi-pythonSkill
- dockerSkill
- frontend-developmentSkill
- humanizerSkill
- monorepoSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
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 14
Read the codebase before making claims.
- Distinctive: line 17
If already in native plan mode, this skill shapes *how* to plan.
- Distinctive: line 21
Check conversation context and skip completed steps.
- Distinctive: line 27
Read relevant files and search the codebase before proposing changes.
- Distinctive: line 35
Don't force every request into the same template.
- Very specific: line 44
Use date +%Y%m%d%H%M%S for timestamps.
≈ “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
ultrathink
Core contract
Figure out the right approach, write a plan the user can inspect, and stop. Planning is not approval to implement — wait for explicit go-ahead before writing any code.
- Clarify only what would materially change the plan.
- Read the codebase before making claims. Distinguish confirmed facts, inferences, and unknowns — never hide uncertainty.
- Optimize for reviewability over momentum.
If already in native plan mode, this skill shapes how to plan; plan mode provides the workflow structure and must follow this skill guidelines.
Process
Check conversation context and skip completed steps.
1. Clarify the planning goal
What kind of plan? Scoped implementation, phased feature, migration/rollout, architecture decision, or spec artifact. Ask only the questions that matter.
2. Ground in evidence
Read relevant files and search the codebase before proposing changes. Be explicit about what's confirmed from code vs. inferred vs. unknown.
3. Choose the right altitude
- Small change → concise plan with touchpoints and validation
- Multi-file feature → phased plan with dependencies
- Architecture choice → options, tradeoffs, recommendation
- Migration / rollout → sequencing, rollback, validation checkpoints
Don't force every request into the same template.
4. Write the plan (right-size the artifact)
Match artifact to plan size — don't force plans/*.md on small scoped changes.
- Inline prose — small single-file tweaks, quick scoped edits, brief decisions. Present the plan in the response and stop.
plans/YYYYMMDDHHMMSS-{plan-name}.md— multi-file features, migrations, architecture decisions, anything that benefits from review or history.plans/YYYYMMDDHHMMSS-{plan-name}/plan.md+ phase files — large multi-phase work where each phase warrants independent reading/editing.
Use date +%Y%m%d%H%M%S for timestamps.
A strong plan covers: problem summary, recommended approach, phases/workstreams, affected files/modules/systems, dependencies and sequencing, validation strategy, risks and mitigations, assumptions and open questions, non-goals when useful.
5. Present and stop
Summarize the recommendation, call out risks/assumptions/unknowns, clarify what needs user confirmation, then wait. Do not drift into coding.
Boundaries
- Prose, not code — describe what changes in prose (name the file, the concept); never include executable syntax (function bodies, JSX, SQL, migration scripts, shell commands). "Illustrative" snippets are still implementation code.
- Right-sized artifacts — don't force heavyweight structure onto small work; don't leave plans so abstract they name no touchpoints or validation path.
- Every plan needs a validation path — how will you know the implementation succeeded?
- Name concrete files, interfaces, and systems where possible. Surface tradeoffs instead of hiding them.
Request
<request>$ARGUMENTS</request>
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.

