What does this skill do, and when should you use it?
This skill helps AI coding agents decompose a specification or clear set of requirements into small, verifiable implementation tasks. It starts with read-only analysis of the specification and relevant code, then maps dependencies, slices work vertically, and defines acceptance criteria, verification steps, dependencies, likely files, and scope for each task. The plan is saved to tasks/plan.md and the checklist to tasks/todo.md. It is intended for work that is large, vague, or dependency-heavy, not obvious single-file changes.
Reads the specification and relevant codebase sections; identifies patterns, conventions, dependencies, risks, and unknowns; maps a dependency graph; divides work into vertical slices; writes tasks containing descriptions, acceptance criteria, verification steps, dependencies, likely files, and scope estimates; orders tasks into phases with checkpoints; saves the implementation plan to tasks/plan.md and the checklist to tasks/todo.md.
- A developer has a specification and needs implementable units before coding.
- A technical lead has a task that feels too large or vague and needs smaller focused sessions.
- Several agents or sessions may work in parallel and need clear dependency boundaries.
- A team needs to communicate scope, verification standards, and open questions to a human reviewer.
How do you install this skill?
- The skill requires writing tasks/plan.md and tasks/todo.md but does not specify confirmation, conflict handling, or rollback before overwriting existing files.
- It references references/definition-of-done.md, which is absent from the supplied evidence; the skill should state which checks cannot proceed when it is missing.
- Repository CI does not establish the quality of this specific skill's planning outputs; users still need to review plans manually.
- No adaptation strategy is stated for Chinese-language tasks, non-Node projects, or mainland-China network constraints.
- Shell / CLI
- Local filesystem
This is an individual skill in the addyosmani/agent-skills repository at skills/planning-and-task-breakdown/SKILL.md. Install the full collection with npx skills add addyosmani/agent-skills. Install only this skill with npx skills add addyosmani/agent-skills --skill planning-and-task-breakdown. The README provides no additional installation procedure specific to this skill.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/addyosmani/agent-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/planning-and-task-breakdown" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Break this specification into implementation tasks. Read the relevant code first, create tasks/plan.md and tasks/todo.md, and do not write code yet.
Place the skill directory where the compatible Agent Skills client expects skills, then provide a specification or clear requirements and ask the agent to plan before coding. Example: "Break this specification into implementation tasks. Read the relevant code first, create tasks/plan.md and tasks/todo.md, and do not write code yet." The workflow expects human review and approval of the plan before implementation begins.
What are this skill's strengths and limitations?
- Requires acceptance criteria and verification steps for every task.
- Makes dependencies, implementation order, risks, and unknowns explicit.
- Promotes vertical slices, checkpoints, and small task sizes.
- Provides reusable templates for a plan document and task checklist.
- It plans and decomposes work but does not implement code, run tests, or deploy software.
- Its expected outputs use the fixed paths tasks/plan.md and tasks/todo.md, which may require adapting an existing project convention.
- The source provides no test suite, runtime metrics, or platform-specific validation for this individual skill.
- Its example verification commands use npm, but the skill does not require Node.js or npm.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Planning & Task Breakdown this page | 54 · Use with care | ★ 107k | 13d ago | MIT |
| Spec-Driven Development | 55 · Use with care | ★ 107k | 13d ago | MIT |
| Harness Long-Running Agent Framework | 50 · Use with care | ★ 2.8k | 5mo ago | AGPL-3.0 |
| Hermes Ephemeral Delegation | 49 · Use with care | ★ 7.8k | 12d ago | MIT |
| Writing Implementation Plans | 42 · Not recommended | ★ 2.9k | 13d ago | Apache-2.0 |
The README compares the overall Agent Skills collection with Superpowers and Matt Pocock's skills, but provides no standalone comparison for this planning skill.
How did FollowSkills review this skill?
The skill requires read-only planning and emphasizes dependencies, risks, acceptance criteria, verification, and human approval. No malware, credential handling, external calls, or destructive defaults are evident; however, data-flow disclosure, sensitive-data handling, rollback guidance, and source attribution are not explicit, so points are deducted.
The workflow, task template, dependency ordering, and checkpoints are largely consistent. Abnormal-input handling and diagnosable stop conditions are thin, and no skill-specific execution tests are supplied; under static calibration the score is capped at 10 and deductions apply.
Audience, use cases, and non-fit cases are clearly stated, including parallelization and task sizing. Input/output boundaries, agent-environment compatibility, Chinese-language support, and mainland-China reachability are not declared, so points are deducted.
The skill has clear structure, progressive workflow, templates, examples, limitations, and a verification checklist. The repository supplies an MIT license, but the skill lacks versioning, changelog, maintenance ownership, and update path; its referenced definition-of-done.md is not included in the supplied evidence, so points are deducted.
It can directly produce tasks/plan.md and tasks/todo.md with acceptance criteria, dependencies, file scope, and verification steps, giving it clear core utility. No representative artifact or independent outcome evidence is provided, and human review remains necessary; static calibration caps this at 7, so points are deducted.
Repository CI provides limited auditable evidence through skill validation, trigger/routing evaluations, and test execution. The supplied evidence does not show coverage of this skill's key paths or independent third-party reproduction, so points are deducted.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →