agentleFS
hyperb1isshyperb1iss/hyperskills/skills/plan/SKILL.md

plan

Skill36 starsChanged 21 days agoMITView on GitHub

Use this skill when decomposing complex work into verifiable tasks and dependencies before or during implementation. Activates on mentions of write a plan, create a plan, break this down, task decomposition, implementation plan, what are the steps, plan the work, spec this out, or decompose this feature.

184 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use heartmula, webhook-subscriptions and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • math-incSkill

    heartmula

    Set up and run HeartMuLa, the open-source music generation model family (Suno-like). Generates full songs from lyrics + tags with multilingual support.

    Also in 9 projects that use this

  • Tommy-ywSkill

    webhook-subscriptions

    Create and manage webhook subscriptions for event-driven agent activation, or for direct push notifications (zero LLM cost). Use when the user wants external services to trigger agent runs OR push notifications to chats.

    Also in 5 projects that use this

  • math-incSkill

    minecraft-modpack-server

    Set up a modded Minecraft server from a CurseForge/Modrinth server pack zip. Covers NeoForge/Forge install, Java version, JVM tuning, firewall, LAN config, backups, and launch scripts.

    Also in 5 projects that use this

  • john-data-chenSkill

    touchdesigner-mcp

    Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.

    Also in 6 projects that use this

  • gaebalaiSkill

    codex-system

    PROACTIVELY consult Codex CLI, your highly capable supporter with exceptional reasoning and task completion abilities. Codex is a trusted expert you should ALWAYS consult BEFORE making decisions on: design choices, implementation approaches, debugging strategies, refactoring plans, or any non-trivial problem. When uncertain, consult Codex. Don't hesitate - Codex provides better analysis. Explicit triggers: "think deeper", "analyze", "second opinion", "consult codex".

    Also in 3 projects that use this

  • ibahgatSkill

    context-optimize

    Improve context efficiency and cache stability for long or complex sessions.

    Also in 3 projects that use this

  • +3 more from math-inc

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 167 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Treat an inherited plan as a hypothesis: recheck consequential claims against the current tree before executing them.

    line 33
  2. Distinctive:

    Keep implementation and its behavioral verification in the same accountable task.

    line 39
  3. Distinctive:

    Use only the fields the task needs:

    line 43

“Common” is measured: kept by at least 6 in 10 of the 167 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Structured Planning

The user's instructions take precedence over this skill; carry existing authorization forward and do not add approval gates. If the user requested only a plan, deliver the plan without starting implementation.

Size the Plan by Uncertainty

Work shapePlanning depth
Clear, reversible change with a direct checkExecute with a brief inline plan
Several dependent behaviors with known patternsRecord tasks, interfaces, and acceptance checks
Uncertain architecture or migrationResolve decisive unknowns and plan a representative vertical slice
Work spanning agents, sessions, or environmentsPreserve ownership, state, dependencies, and resumption steps

File count is a poor proxy for risk: a one-line authorization change can need more analysis than a mechanical rename across many files.

Establish the Actual Starting Point

Inspect repository status and relevant diffs, and locate the real verification commands in scripts, hooks, or CI. Record the facts that constrain execution:

  • The observable user or system outcome.
  • The baseline revision or environment inspected.
  • The ship state (pre-launch, single consumer, or production with users), which decides how much migration machinery the plan may carry.
  • Affected interfaces, ownership, and compatibility obligations.
  • Uncertainty that could invalidate the work.
  • Explicit constraints, non-goals, and user corrections, with requirements ("must") kept distinct from preferences stated casually.
  • Constraints the user treats as obvious and left unsaid: a convention, a taste, an integration they assume. Where two readings produce materially different work, ask once with the pick stated; otherwise record the reading taken.

Treat an inherited plan as a hypothesis: recheck consequential claims against the current tree before executing them. When a premise fails, update the plan itself, because a detached comment leaves the executor on stale instructions.

Choose a Useful Slice

Prefer a vertical slice that carries a real input through the uncertain boundary to the consumer that needs it. Simplification must preserve the requested result: do not replace a difficult product goal with a convenient demonstration of plumbing. Sequence migrations by compatibility and deployment; source-file order does not determine safe rollout order.

Keep implementation and its behavioral verification in the same accountable task. A title containing "and" is not evidence that the task should split; split at independently reviewable outcomes, distinct owners, or real dependency boundaries.

Give Each Task an Executable Contract

Use only the fields the task needs:

FieldContent
OutcomeBehavior or artifact delivered
ScopeKnown files/modules and permitted expansion rules
DependenciesRequired inputs or interfaces, with their producer
VerificationExact runnable check plus the behavior it proves
Completion evidenceExpected artifact, assertion, observation, or report
Risk and recoveryCompatibility, rollback, or unresolved external dependency when material
### Task: Reject expired sessions at the request boundary

Outcome: An expired session receives the established unauthorized response.
Scope: The session validator and its request-level tests.
Depends on: The existing clock injection interface, confirmed in the repo.
Verify: [repository's exact focused test command].
Cases: Expired, valid, and boundary-time sessions preserve expected behavior.
Completion: Tests exercise the request path and report executed cases.

A copyable command still needs an assertion: "renders" does not prove that the generated policy matches the intended resources.

Map Dependencies and Ownership

File separation does not prove independence: two modules can share a schema, generated file, test database, port, or output directory. Name those shared resources and their owners. Designate an integration owner who verifies combined behavior after branches or patches meet.

Make the Plan Resumable

Select one authoritative plan and link it from the project's tracker rather than duplicating it across chat, repository files, and memory. Do not create or commit planning documents merely because work spans sessions. A plan is read by a human and a future session, not by the model that wrote it: keep it to the decisions, contracts, and open calls, and when the host renders rich artifacts, put each open question beside the diagram or snippet it concerns.

A checkpoint records the current revision, completed outcomes and evidence, active ownership, pending decisions, and the exact next step. On resumption, re-read the checkpoint and inspect live state; notes describe a past state.

Replan Without Losing the Goal

When implementation exposes a new constraint, carry the correction through tables and acceptance criteria, not just the edited paragraph. A case the plan did not cover gets its call recorded beside the task it affects, with the alternative rejected; a call that lives only in the code is lost at the next checkpoint. At integration boundaries, compare the accumulated diff with the intended outcome; tests can pass while scope drifts.

Report an unavailable external gate separately from completed local checks, with its blocker and next action.

Evidence

Reviewed 2026-09-04 and 2026-10-27; sources and their limits are in references/evidence.md.

Anti-Patterns

Anti-patternBetter move
Exact edit predictions before reading codeSpecify verified boundaries, mark uncertainty
Separate every feature from its testsKeep one owner accountable for the behavior
Parallelize based only on filenamesInspect shared interfaces and runtime state
Ease review because the task is lateMatch review to current risk and evidence
Keep decomposing after success is coveredExecute; replan when evidence changes the work

What This Skill is NOT

  • A requirement for trivial work or a fixed file-count threshold.
  • Permission to execute a plan-only request.
  • A promise that an initial architecture, schedule, or task list cannot change.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.