space

Documents the sprint-end refinement session: promotes WP-local ADR candidates into solution.md, archives superseded work-package design sections, and produces a refine-session.md recording every decision made. Use when the user mentions "refine docs", "promote ADRs", or "end of sprint cleanup". Do NOT use before implementation is complete — use implement first. Do NOT use to review docs before a sprint — use review-docs for that.

Documents a consequential architecture decision as an ADR-NNNN.md file. Use when the user mentions "write an ADR", "document this architecture decision", "ADR for {choice}", or "record the decision about {topic}". Covers the context, options considered, decision, and consequences. Do NOT use to identify which decisions need ADRs — use plan-adr first. Do NOT use to review a draft ADR — use review-adr.

Drafts a domain-level or work-package backlog.md. Use when the user mentions "backlog", "epic list", "stories", "decompose", or "write the backlog for {domain}". Domain scope defaults to the Now phase only — use --depth full for all phases. Work-package scope produces EARS + Gherkin acceptance criteria per the canonical story schema. Do NOT use for solution architecture — use write-solution. Do NOT use for roadmaps — use write-roadmap.

Drafts product.md at portfolio, product, or domain scope. Portfolio scope binds multiple products with thesis, sequencing, and commercial model. Product scope produces a pitch (Phase 0, ≤2 pages) or full strategy (Phase 2+, ≤5 pages). Use when the user mentions "product doc", "PRD", "product strategy", or "what are we building". Do NOT include tech stack — use write-solution. Do NOT use for roadmaps — use write-roadmap. Do NOT review an existing strategy — use review-product.

Drafts solution.md in stub mode (Phase 0, two sections) or full arc42-lite mode (Phase 2+, ten sections). Use when the user mentions "solution design", "architecture", "how will this work", or "what should we build". Stub mode produces a Phase 0 anchor with [NEEDS CLARIFICATION] placeholders; full mode reverse-engineers the architecture from a shipped walking skeleton. Do NOT use for business strategy — use write-product. Do NOT use for sprint TDD — use write-wp-design.

Drafts tech-stack.md defining the technology choices for a product or domain, with rationale and trade-offs for each decision. Use when the user mentions "define the tech stack", "document technology choices", "tech stack for {product}", or "what technology should we use". Do NOT use for full solution architecture — use write-solution. Do NOT use for individual technical decisions — use write-adr for consequential choices.

Documents the sprint-end refinement session: promotes WP-local ADR candidates into solution.md, archives superseded work-package design sections, and produces a refine-session.md recording every decision made. Use when the user mentions "refine docs", "promote ADRs", or "end of sprint cleanup". Do NOT use before implementation is complete — use implement first. Do NOT use to review docs before a sprint — use review-docs for that.

Performs a comprehensive, first-principles audit of a codebase as a Senior Software Architect. Use when the user mentions "review the codebase", "audit the repo", "check the overall code quality", or "give me a health check on {repo}". Produces a prioritised finding list with file-level evidence. Do NOT use for a single PR diff — use review-code for that. Do NOT use for design documents — use review-solution for that.

Documents a consequential architecture decision as an ADR-NNNN.md file. Use when the user mentions "write an ADR", "document this architecture decision", "ADR for {choice}", or "record the decision about {topic}". Covers the context, options considered, decision, and consequences. Do NOT use to identify which decisions need ADRs — use plan-adr first. Do NOT use to review a draft ADR — use review-adr.

Drafts a domain-level or work-package backlog.md. Use when the user mentions "backlog", "epic list", "stories", "decompose", or "write the backlog for {domain}". Domain scope defaults to the Now phase only — use --depth full for all phases. Work-package scope produces EARS + Gherkin acceptance criteria per the canonical story schema. Do NOT use for solution architecture — use write-solution. Do NOT use for roadmaps — use write-roadmap.

Drafts product.md at portfolio, product, or domain scope. Portfolio scope binds multiple products with thesis, sequencing, and commercial model. Product scope produces a pitch (Phase 0, ≤2 pages) or full strategy (Phase 2+, ≤5 pages). Use when the user mentions "product doc", "PRD", "product strategy", or "what are we building". Do NOT include tech stack — use write-solution. Do NOT use for roadmaps — use write-roadmap. Do NOT review an existing strategy — use review-product.

Drafts solution.md in stub mode (Phase 0, two sections) or full arc42-lite mode (Phase 2+, ten sections). Use when the user mentions "solution design", "architecture", "how will this work", or "what should we build". Stub mode produces a Phase 0 anchor with [NEEDS CLARIFICATION] placeholders; full mode reverse-engineers the architecture from a shipped walking skeleton. Do NOT use for business strategy — use write-product. Do NOT use for sprint TDD — use write-wp-design.

Identifies the right skill for a vague or open-ended request. Use when the user asks "which skill should I use?", "what can I do?", or "how do I start this?". Presents a CI-generated index of all stable skills with names, artefacts, phases, roles, and trigger conditions so the agent can route to the correct skill. Do NOT use to produce a delivery artefact -- use write-product, write-solution, write-backlog, or write-wp-design for that. Do NOT use to implement a story -- use implement.

Implements code for a story or task against an approved design.md and backlog.md. Use when the user mentions "implement", "build", "code this story", "write the code for {story}", or "make this feature work". Reads the design document first, then makes targeted changes following existing patterns. Do NOT use before design is approved — use write-wp-design first. Do NOT use for code review — use review-code after implementation.

Performs targeted code refactoring to address issues raised in a code review or to improve quality without changing behaviour. Use when the user mentions "refactor", "fix the review comments", "address the feedback", "clean this up", or "fix the issues from the code review". Do NOT use to add new features — use implement for that. Do NOT use to review code — use review-code for that.

Performs a comprehensive code review of changes in a branch or PR. Use when the user mentions "review this code", "review the PR", "check the diff for {story}", or "code review feat/{branch}". Checks quality, correctness, security, and compliance with design.md and backlog.md acceptance criteria. Do NOT use before implementation is done — use implement first. Do NOT use to review design documents — use review-docs for that.

Refines backlog.md by applying five grooming activities: prioritise, break down, estimate, define acceptance criteria, and remove. Works at domain (epic) or work-package (story) scope. Use when the user mentions "refine the backlog", "groom the backlog", "re-prioritise the epics", "the stories need AC", or "the backlog is stale". Do NOT use to write a backlog from scratch — use write-backlog. Do NOT use for a critical quality review — use review-backlog.

Refines product.md on a regular cadence by recording sprint learnings, updating metric baselines, closing resolved open questions, and advancing scope status. Works at portfolio, product, or domain scope. Use when the user mentions "update the product strategy after this sprint", "refine product.md", "the product doc is stale", or "record what we learned". Do NOT use for a critical quality review — use review-product. Do NOT use to write from scratch — use write-product.

Refines roadmap.md to reflect delivery reality — advancing phase status, recording exit-criteria evidence, updating cross-dependency statuses, and adjusting phase contents when sprint outcomes diverge from plan. Use when the user mentions "update the roadmap after this sprint", "advance the phase", or "the roadmap is out of date". Do NOT use for a critical credibility review — use review-roadmap. Do NOT use to write from scratch — use write-roadmap.

Refines solution.md after a sprint or phase to reflect what was built — updating the building-block view, runtime sequences, data model, risk register, ADR log, and emergent patterns. Use when the user mentions "update the solution architecture", "the solution.md has drifted", or "refine the architecture after this sprint". Do NOT write from scratch — use write-solution. Do NOT conduct a quality review — use review-solution.

Reviews backlog.md at domain or work-package scope as a Senior Delivery Lead, checking strategic alignment, AC completeness, dependency integrity, and sprint feasibility. Amends the backlog and appends a verdict. Use when the user mentions "review the backlog", "is this backlog ready", "backlog quality review for {name}", or "are these stories ready". Do NOT use for incremental grooming — use refine-backlog. Do NOT use to write a backlog — use write-backlog.

Reviews a work-package design.md for implementation readiness — checking that the design is implementable, APIs are fully specified, error handling covers all failure modes, and the test strategy is defined. Amends the document and appends a verdict. Use when the user mentions "review the design for {epic}", "is the WP design ready", or "check the design before the sprint". Do NOT use to review product strategy — use review-product. Do NOT use to review code — use review-code.

Reviews product.md and solution.md for completeness and alignment before development begins. Use when the user mentions "review the docs for {epic}", "check the design and requirements are ready", or "are these docs good enough to start building". Flags gaps, inconsistencies, and drift between the two documents. Do NOT use to review code — use review-code for that. Do NOT use to review an ADR — use review-adr.

Reviews product.md at portfolio, product, or domain scope as a critical Senior Product Manager — checking strategy coherence, problem specificity, scope discipline, and commercial validity. Amends the document and appends a verdict. Use when the user mentions "review the product strategy", "critique product.md", or "is this strategy any good". Do NOT use to write a strategy — use write-product. Do NOT use to review implementation design — use review-design.

Reviews roadmap.md at portfolio, product, or domain scope as a Senior Delivery Lead — checking phase coherence, exit-criteria testability, sequencing rationale, feasibility, and dependency completeness. Amends and appends a verdict. Use when the user mentions "review the roadmap", "is this roadmap credible", or "critique the delivery plan". Do NOT use for post-sprint updates — use refine-roadmap. Do NOT use to write a roadmap — use write-roadmap.

Reviews solution.md as a Senior Solution Architect, checking structural soundness, section completeness, NFR specificity, ADR coverage, and drift from what was actually built. Amends the document and appends a verdict. Use when the user mentions "review the architecture", "is the solution sound", or "review solution.md for {name}". Do NOT use to update solution.md incrementally — use refine-solution. Do NOT use to review product strategy — use review-product.

Identifies the architecture decisions that need ADRs and produces a prioritised adr-plan.md before technical design proceeds. Use when the user mentions "plan ADRs", "what ADRs do we need", "identify architecture decisions for {domain}", or "which decisions need documenting". Produces a list of proposed ADRs — does not write full ADRs. Do NOT use to write an ADR — use write-adr. Do NOT use to review an existing ADR — use review-adr.

Produces contracts.md for a domain as an executable index of types, Zod schemas, API route contracts, and analytics event payloads. Use when the user asks for "contracts", "types", "schema", or "interface definitions" for a domain. Output is TypeScript / Zod / OpenAPI source with one worked example per contract, not prose. Do NOT use for solution architecture — use write-solution. Do NOT use for work-package design — use write-wp-design.

Drafts roadmap.md at portfolio, product, or domain scope using outcome-based phases with exit criteria, not epic lists. Portfolio scope sequences multiple products; product scope phases a single product; domain scope phases a bounded context. Use when the user mentions "write the roadmap for {name}", "create the delivery roadmap", or "sequence the phases". Do NOT use to list epics — use write-backlog. Do NOT use before product.md exists — use write-product first.

Drafts a work-package design.md in walking-skeleton mode (foundation sprint, 2–4 pages) or TDD mode (sprint 2+, 5–10 pages). Use when the user mentions "design", "TDD", "technical design", "write the design for {epic}", or "how should we implement {story}". Walking-skeleton names the slice, files, and acceptance gates. TDD adds runtime flows, exact signatures, and cross-squad coordination. Do NOT re-narrate solution.md patterns — cite them. Do NOT include business context — use write-product.

Creates a merge request or pull request for the current branch after implementation is complete. Use when the user mentions "create an MR", "open a PR", "raise a pull request", or "merge this branch". Generates a title, description, labels, and reviewer suggestions from the diff and any linked story. Do NOT use for code review — use review-code. Do NOT use before implementation is done — use implement first.

Alias for write-wp-design. Use write-wp-design instead. This alias will be removed in @daddia/skills v0.4.0. write-wp-design drafts a work-package design document in walking-skeleton mode (foundation sprint, 2–4 pages) or TDD mode (sprint 2+, 5–10 pages). Use when the user mentions "design", "TDD", "write the design for {epic}", or "how should we implement {story}".

Implements code for a story or task against an approved design.md and backlog.md. Use when the user mentions "implement", "build", "code this story", "write the code for {story}", or "make this feature work". Reads the design document first, then makes targeted changes following existing patterns. Do NOT use before design is approved — use write-wp-design first. Do NOT use for code review — use review-code after implementation.

Identifies the architecture decisions that need ADRs and produces a prioritised adr-plan.md before technical design proceeds. Use when the user mentions "plan ADRs", "what ADRs do we need", "identify architecture decisions for {domain}", or "which decisions need documenting". Produces a list of proposed ADRs — does not write full ADRs. Do NOT use to write an ADR — use write-adr. Do NOT use to review an existing ADR — use review-adr.

Produces a delivery-plan.md that sequences the Phase-0 artefacts for a new portfolio, product, or domain before the foundation sprint starts. Use when the user mentions "plan delivery for {name}", "sequence the artefacts", or "how do I start this?". Do NOT use to author any individual artefact — use write-product, write-solution, write-roadmap, write-backlog, or write-contracts for that.

Performs targeted code refactoring to address issues raised in a code review or to improve quality without changing behaviour. Use when the user mentions "refactor", "fix the review comments", "address the feedback", "clean this up", or "fix the issues from the code review". Do NOT use to add new features — use implement for that. Do NOT use to review code — use review-code for that.

Refines backlog.md by applying five grooming activities: prioritise, break down, estimate, define acceptance criteria, and remove. Works at domain (epic) or work-package (story) scope. Use when the user mentions "refine the backlog", "groom the backlog", "re-prioritise the epics", "the stories need AC", or "the backlog is stale". Do NOT use to write a backlog from scratch — use write-backlog. Do NOT use for a critical quality review — use review-backlog.

Refines product.md on a regular cadence by recording sprint learnings, updating metric baselines, closing resolved open questions, and advancing scope status. Works at portfolio, product, or domain scope. Use when the user mentions "update the product strategy after this sprint", "refine product.md", "the product doc is stale", or "record what we learned". Do NOT use for a critical quality review — use review-product. Do NOT use to write from scratch — use write-product.

Refines roadmap.md to reflect delivery reality — advancing phase status, recording exit-criteria evidence, updating cross-dependency statuses, and adjusting phase contents when sprint outcomes diverge from plan. Use when the user mentions "update the roadmap after this sprint", "advance the phase", or "the roadmap is out of date". Do NOT use for a critical credibility review — use review-roadmap. Do NOT use to write from scratch — use write-roadmap.

Refines solution.md after a sprint or phase to reflect what was built — updating the building-block view, runtime sequences, data model, risk register, ADR log, and emergent patterns. Use when the user mentions "update the solution architecture", "the solution.md has drifted", or "refine the architecture after this sprint". Do NOT write from scratch — use write-solution. Do NOT conduct a quality review — use review-solution.