A collection of skills for product development workflows by Unlearn.
npx skills add unlearndev/skillsThis uses skills.sh to install all skills into your project's .claude/skills/ directory. Once installed, your agent will automatically detect and use them based on your prompts.
To install a single skill:
npx skills add unlearndev/skills --skill spec-generator| Skill | Description |
|---|---|
| spec-generator | Turn a vague idea into a detailed product spec |
| feature-generator | Expand a spec into an implementation-ready feature list |
| code-review | Review staged changes or a specific area of the codebase |
| checklist | Convert plans or reviews into persistent markdown checklists |
| review-order | Prepare a structured, scannable review map of a branch's changes |
| first-five | Triage a branch against the First Five checklist and flag only real concerns |
| triage | Group a branch's changes into feature areas and assign risk tiers |
| zombies | Suggest the most relevant tests to write for a feature using the ZOMBIES heuristic |
| warm | Evaluate newly added or upgraded dependencies against the WARM check |
| preflight | Produce a production pre-flight checklist of everything a branch needs once it merges |
| ship-gate | Scan the staged diff for secrets, debug code, and stray files and return a straight PASS or BLOCK |
| qa-checklist | Turn a branch's diff into a short, click-through manual QA checklist for the feature |
| judge-arch | Judge a branch's changes against the design decisions in the repo's DECISIONS.md |
Generate a detailed product specification from a rough idea, description, or supporting materials (screenshots, notes, wireframes).
> Write up a spec for a Slack bot that summarizes daily standups
> Turn this idea into a requirements doc
> /spec-generator
Outputs a structured markdown spec covering overview, user stories, data model, API, UI, and edge cases.
Expand a spec.md into a full features.md ordered by implementation dependency, or keep the two files in sync when either changes.
> Generate features from the spec
> Sync spec and features
> /feature-generator
Features are ordered for optimal AI agent implementation: foundation first, then simple flows, complex features, admin features, and cross-cutting concerns.
Review staged git changes or a specific area of the codebase. Optionally delegate to an alternative agent (codex, aider, goose).
> Review my staged changes
> /code-review codex auth-module
> /code-review
Findings are classified by severity: Critical, Error, Warning, Suggestion, and Nitpick.
Convert the current plan, code review, or any structured content into a persistent markdown checklist saved to .claude/plans/.
> Turn this plan into a checklist
> /checklist
Creates a trackable file with checkboxes that the agent checks off as it works through the implementation.
Prepare a scannable map of a branch's changes, grouped by feature and ordered for review (Types → Data Flow → Business Logic → Edge Cases) with file:line citations.
> Prep a review of this branch
> /review-order
> /review-order develop
Outputs a descriptive map (no suggestions, no questions) so you can jump straight into the code yourself.
Scan a branch or diff against the First Five checklist (Error Handling, Input Boundaries, External Calls, State Mutations, Assumed Dependencies) and report only genuine concerns — verified against the codebase, not flagged on suspicion.
> Run the first five on this branch
> /first-five
> /first-five develop
Outputs a short, scannable triage list with file:line citations and
Map a branch's changes into feature-area groups, each tiered as High / Medium / Low risk, so you can decide where to spend your review time. This is a triage map — not a review.
> Triage this branch
> /triage
> /triage develop
Outputs a risk-ordered report grouped by feature area (e.g. "Authentication", "Notification Delivery"), with auto-generated files pushed to a Skip section. No suggestions, no fixes — just a map.
Identify the most valuable tests to write for a feature using the ZOMBIES heuristic (Zero, One, Many, Boundaries, Interface, Exceptions, Simple scenarios). Pass a free-text feature description, or omit the argument to use the current branch's diff.
> /zombies
> /zombies sign-in code login flow
> /zombies image upload validation
Outputs a grouped list of test ideas — only the ZOMBIES categories that genuinely apply, with specific values pulled from the code (column lengths, expiry windows, throttle limits) so you can see at a glance what's worth covering and write the tests yourself.
Evaluate newly added or upgraded dependencies against the WARM check (Worth it, Alive, Right-sized, Maintained securely). Covers client- and server-side dependencies in any language by diffing the changed manifests against a base.
> WARM check this branch
> /warm
> /warm develop
Outputs a per-dependency report scoring each WARM letter with a one-word verdict (Keep, Reconsider, Patch/Pin, Replace), ordered by concern. Facts come from real registry, repo, and advisory lookups — nothing fabricated. Evaluates and reports only; it never edits manifests or runs installs.
Read the diff between the current branch and main and produce a production pre-flight checklist of everything that must be done or configured once the change merges (migrations, queue workers, env vars, new services).
> What needs doing in production before this merges?
> /preflight
Outputs a short checklist grouped by type (Database, Infrastructure, Configuration, Operational), sub-grouped by confidence, with a file citation for every item.
Scan the staged diff for the four things that should never be committed — secrets or API keys, leftover debug code (dd(), console.log), large commented-out blocks, and accidentally added large or binary files — reporting the file and line for each.
> Check what's staged before I commit
> Anything I shouldn't commit here?
> /ship-gate
Ends with a single PASS or BLOCK verdict, so it drops cleanly into a pre-commit or pre-push hook. Reports only — it never unstages or edits anything.
Read a branch's diff and produce a short, manual QA checklist to run against the feature — only the behaviour this change touched, written as plain-English steps a person can click through.
> What should I test on this branch after deploy?
> /qa-checklist
> /qa-checklist develop
Outputs a ~5–10 item markdown checklist, each step pairing an action with its expected result, with any required preconditions (test accounts, feature flags) noted up front.
Judge the current branch's changes against the design decisions recorded in the repo's DECISIONS.md — only those decisions, no other findings or style feedback. Every violation needs evidence you can point at in the change: a file and line, a whole file, or the set of files involved.
> Judge this branch against our decisions
> Check this change against DECISIONS.md
> /judge-arch
Outputs a report grouped by violated decision, each violation citing where it is, what the code does, and the smallest concrete fix that satisfies the decision.