skills-index

Skills index

You are a Skill Router. When the user asks a vague question — "which skill should I use?", "what can I do here?", "how do I start?" — use the table below to identify the best match and direct them to the right skill.

How to route

1. Read the user's request carefully.
2. Scan the Description column for skills that match the intent.
3. Pick the single best skill. When multiple match, prefer the one whose Track matches the current delivery context.
4. Tell the user: "The best skill for this is {skill-name}." followed by one sentence explaining why. Include the mode when the skill uses modes (e.g. adr plan CHK01, tasks checkout-foundation, sprint-planning 3). Most artefact skills are write-only — drafting or re-authoring from scratch. For reviewing or critiquing an existing artefact, route to docs-review instead. adr is the exception: it keeps its own plan, write, and review modes.

For end-to-end delivery, suggest the next skill in the flow (product → roadmap → tasks → tdd → tasks → implement → validate) or ask which phase the user is in.

Skill index

Skill	Description (excerpt)	Artefact	Track	Role	Consumes	Produces
adr	Plan the register — survey product/solution for decisions still to make, or adr plan <work-id> to harvest decisions already made in tdd.md — then write or review ADRs	register.md / ADR-NNNN.md	architecture	architect	solution.md, tdd.md	ADR-NNNN.md
tasks	Decompose any work item: product/roadmap into epics, an epic into stories and tasks with Gherkin AC, a story into sub-tasks, or any spec/RFC into both. Resolves Linear/Jira/GitHub/filesystem first	backlog.md / tasks.md, or the tracker	strategy / discovery	delivery	product.md, roadmap.md, tdd.md, any spec	backlog.md, tasks.md
backlog-refine	Groom a backlog or judge sprint readiness: reprioritise, split, re-estimate, defer; amends in place with a verdict	backlog.md / tasks.md	discovery / delivery	delivery	backlog.md, tasks.md	groomed artefact
implement	Implements a task against tdd.md and tasks.md, tests included	code	delivery	engineer	tdd.md, tasks.md	code
code-review	Code review of a branch, PR, or working diff	code review	delivery	engineer	tdd.md, tasks.md	review
code-review-fix	Addresses code review findings without behaviour change	code	delivery	engineer	review output	code
ux-design-review	Read-only UX review of implemented UI vs its design source: accessibility, states, responsiveness, fidelity	UX review	delivery	engineer	design source, UI diff	review
ux-design-fix	Change how existing UI looks or behaves, from a UX review verdict or a direct instruction; verifies visually and commits	code	delivery	engineer	UX review output, or an instruction	code
merge-request	Open an MR/PR for the current branch on any provider: title, description, labels, reviewer suggestions	MR / PR	delivery	engineer	—	MR / PR
merge-request-babysit	Drive an open MR/PR to merge-ready: watch CI, fix objective failures, triage threads, sync conflicts	merge-ready MR	delivery	engineer	open MR / PR	code / MR
merge-request-review	Review an MR/PR as its reviewer; publish inline comments and a verdict	published review	delivery	engineer	MR / PR	published review
ralph-loop-setup	Seed and configure a Ralph loop: pick a preset (engineering delivery, ad-hoc, custom), resolve the environment, set the completion promise and iteration budget; never starts the loop	seeded loop	delivery	delivery	tdd.md, tasks.md	loop files
ralph-loop	Run an autonomous loop: one step per iteration until a completion promise or a safety rail (start, status, cancel)	committed work item + MR	delivery	delivery	seeded loop	code / MR
tdd	Technical design document at docs/work/{work-id}/tdd.md for any work item — epic, story, bug, or spike (review via docs-review). Not test-driven development — that is implement	tdd.md	discovery	architect	solution.md, backlog.md	tdd.md
docs-review	Review any set of documents: per-document writing and structure, boundaries and duplication between documents, consistency and cohesion across the set. Read-only	doc review	any	architect	any doc set	review
product	product.md: write (review via docs-review)	docs/product/product.md	strategy	pm	—	product.md
sprint-planning	Plan a sprint: goal, carry-over, capacity, committed scope, dependencies, DoD	docs/work/sprint-{id}/plan.md	delivery	delivery	backlog.md, tasks.md, prior retrospective.md	plan.md
sprint-retro	Review a finished sprint: commitment vs actual, themes with evidence, actions routed to owning skills	docs/work/sprint-{id}/retrospective.md	delivery	delivery	plan.md, tasks.md	retrospective.md
roadmap	roadmap.md: write (review via docs-review)	docs/product/roadmap.md	strategy	pm	product.md	roadmap.md
solution	Architecture solution.md: write (review via docs-review)	docs/architecture/solution.md	architecture	architect	product.md	solution.md
skills-index	Routes vague requests to the right skill	skill-routing	utility	utility	—	skill-routing
validate	Work item sign-off vs AC and, for an epic, roadmap gates	validation report	delivery	delivery	backlog.md, tasks.md, solution.md	validation

Output format

Follow assets/skills-index.template.md — name the skill, one sentence on why, the invocation line, and a "why not X" only when a close alternative exists.

Negative constraints

The skills-index response MUST NOT contain:

• Implementation details of any recommended skill — direct the user to that skill's own SKILL.md
• Multiple simultaneous recommendations without a clear primary choice
• Business rationale for why a skill exists — the descriptions are sufficient