agentsclimarketplace

One initiative brief

Skill onehorizonai/skills/skills/one-initiative-brief

Draft a structured initiative brief for roadmap-first planned work in One Horizon. Use for "write an initiative", "draft an initiative brief", "plan this initiative", or "scope this roadmap work". Requires One Horizon MCP.From its SKILL.md

Install
npx -y skills add onehorizonai/skills --skill one-initiative-brief

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

9.0 KB, as published. Nobody here has run it

Initiative Brief

Turn a rough roadmap idea into a clear initiative brief, then create it or finalize an existing draft.

Core rule

  • Understand the problem before proposing solutions; produce a design doc, not code
  • Decide mode early and keep it stable:
    • new initiative → flow ends in create
    • existing initiative draft → flow ends in finalize/update, not create
  • Write in markdown; use tables or Mermaid only when they clarify the design
  • Preserve existing media in canonical markdown exactly (![alt](url), [video](url#one-video=1), [youtube](url#one-youtube=1), [figma](url#one-figma=1)) — do not remove, replace, or normalize unless the user asks
  • Edit existing descriptions with patch-document + initiative taskId; use update-initiative for metadata only
  • Product perspective only — not implementation, engineering tasks, or estimates
  • Default to feature-level scoping unless the user describes broader product/company work
  • Business goals are supporting context, not the backbone
  • Acceptance criteria: observable, user-verifiable ("user can do X on Y screen")
  • Name user-facing surfaces in feature sections
  • Out-of-scope items need a short reason so boundaries hold without chat context

Boundaries

Use when:

  • shaping roadmap-first planned work
  • initiative needs background, scope, non-goals, risk, or rollout clarity
  • user needs a brief others can review and execute from

Do not use when:

  • brief is complete and user only wants the record created
  • request is a bug, ongoing work, or Todo
  • user wants a status update, not a new brief

Other skills:

  • one-create-task — initiative already clear enough to create directly
  • one-task-management — operational lookup, assignment, tagging outside this flow
  • one-initiative-summary — reporting on existing initiatives

Metadata

  • Pull taxonomy before creation when product, customer, company, goal, or component signals exist
  • Prefer product labels first; attach goals, components, company/customer labels on high-confidence matches only
  • list-taxonomy only after scope is stable
  • Multiple plausible labels → ask, don't guess
  • Clear parent roadmap effort → set parentInitiativeId
  • Owner/parent in structured metadata — not Owner: lines in markdown unless requested
  • Related work → URLs or markdown links, not plain text labels

Conversation

  • Open with what they want to build, who it is for, constraints, and ideas they already have
  • Interview relentlessly — one question at a time, reuse prior answers — until you have shared understanding; do not draft early
  • "Just do it" or fully formed plan → still cover missing design-discovery items and premise challenge before drafting
  • Company mode (customers, revenue, GTM) → harder evidence-driven questions
  • Take a position during discovery — no filler hedging

Execution order

  1. Confirm initiative-shaped work (not bug, feature request, or Todo)
  2. Set mode: existing initiative/task ID/draft → existing initiative draft; else → new initiative
  3. Opening intake + design discovery until shared understanding (or N/A for non-interface work)
  4. Check related initiatives and parent linkage
  5. Resolve taxonomy after scope is stable
  6. Draft brief → resolve metadata gaps → after approval, follow mode end state

Minimum brief

Draft only after design discovery is complete (or explicitly N/A for non-interface work) and premises are agreed.

Required before draft:

  • user/workflow, JTBD, what they are trying to accomplish
  • problem and solution direction (solution can be partial or TBD)
  • what changes this phase; in/out scope
  • what success looks like — observable, user-verifiable
  • product (if known)

Put remaining uncertainty in ### Open questions.

Output

  • Markdown brief, no H1; TLDR paragraph first; ### major sections, #### sub-sections
  • Tables for tradeoffs/owners/phases; Mermaid for flows/rollout — only when they help
  • Related work as URLs or markdown links
  • ### Invariants when behaviors or contracts must not break

Phase 1: Context

  1. list-initiatives (active statuses) + list-completed-work for relevant team/workspace
  2. Opening intake (one question at a time, skip what's already answered):
    • What do you want to build?
    • Who is it for?
    • Constraints or requirements already fixed?
    • Ideas or directions you already have?
  3. Product stage only when adoption changes scope/evidence/rollout (pre-product / has users / has paying customers); skip for internal or website content
  4. Background only if unclear: why now, what's not good enough today
  5. Scoping questions until answered: who, JTBD, what they are trying to accomplish, in/out scope, smallest useful version, proof of success, short business note
  6. Missing taxonomy: product area, customer/company/segment tag

Design discovery

Required for user-facing interfaces, screens, or flows. Skip only when work has no interface surface — note "N/A — no user-facing interface" in the brief.

Cover every item below before drafting. Keep asking until each is answered or explicitly deferred:

  • Primary user, JTBD, what they are trying to accomplish
  • What success looks like for this interface
  • Hard constraints: devices, accessibility, performance budgets, brand guidelines
  • Content the interface will contain; what is placeholder vs real

Phase 2: Related work

  1. Extract 3-5 keywords after user states the problem
  2. search-tasks with categories: ["initiative"]; get-task-details on hits
  3. Strong overlap → FYI: Related initiative found: [{title}](<url>). Key overlap: {one-line}. Ask build on prior or start fresh
  4. No match → proceed silently
  5. Clear parent → propose parentInitiativeId

Phase 3: Landscape (optional)

  • Ask consent before external search (generalized terms only, not proprietary/stealth names)
  • Skip if unavailable or declined
  • Read 2-3 results; note standard approach vs what this conversation suggests
  • Useful insight → state plainly; otherwise build on the standard path

Phase 4: Premise challenge

Get agreement before solutions:

  • Right problem? What if we do nothing?
  • Existing workflows that partially solve this?
  • Scoped as feature/phase, not full product?
  • In/out boundaries crisp?
  • What must stay true for the user?
  • If users/paying customers exist, does evidence support this?
PREMISES:
1. <statement>: agree or disagree?
2. <statement>: agree or disagree?
3. <statement>: agree or disagree?

Disagree → revise and loop.

Phase 5: Brief template

Keep one canonical markdown brief updated through the session.

TLDR (2-4 sentences): initiative, user/workflow, this phase, main boundary

### Problem
Who is affected, what is broken or missing today, why it matters now

### Solution
Proposed direction for this phase — partial or TBD is fine; say what is decided vs still open

### Background
What is changing, why now, cost of inaction. Business reason brief and secondary.

### Design (user-facing work only)
Primary user and JTBD; success for this interface; hard constraints; content (real vs placeholder). Omit if N/A.

### Feature / use case sections
Concrete titles like `### Add Login with Google`. Short paragraph each: who, what changes, why. Name screen/entry point when helpful. No `### User story` heading.

### In scope
Behaviors, surfaces, flows, constraints for this phase

### Out of scope
`- Item — excluded because <reason>`. Related ideas not pulled in.

### Acceptance criteria
Observable user-verifiable statements — not metrics. One end-to-end verification step.

### Invariants
What must remain true; omit if nothing specific

### Assumptions, risks, and open questions
Assumptions, blockers, open decisions

### Rollout / handoff
Pilot vs first release vs full rollout; who to inform; post-launch owner

Finalize

After user approves:

  1. Resolve owner, team, taxonomy, parent via one-find-team and list-taxonomy
  2. Brief body: background, scope, boundaries, risks, rollout only
  3. new initiative → confirm title, brief, workspace, owner/team, taxonomy → create-initiative
  4. existing initiative draft → patch-document for description; update-initiative for metadata only
  5. Closing question matches mode; never ask "ready to create?" in existing initiative draft mode
create-initiative({
  "title": "<initiative name>",
  "description": "<full initiative brief in markdown>",
  "status": "Open",
  "workspaceId": "<workspaceId>",
  "assigneeIds": ["<userId>"],
  "teamIds": ["<teamId>"],
  "parentInitiativeId": "<parentInitiativeId>",
  "taxonomyLabelIds": ["<productLabelId>", "<customerOrCompanyLabelId>"]
})

Post-creation revisions: patch-document with precise ops (replace_text, insert_before, insert_after, delete_text); update-initiative for metadata only.

Keep looking

Skills are one crate of 609,689. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.