AgentStack
SKILL verified MIT Self-run

One Roadmap Suggestions

skill-onehorizonai-skills-one-roadmap-suggestions · by onehorizonai

Suggest One Horizon roadmap improvements by reviewing initiatives, planned work, blockers, bugs, and taxonomy. Use when asked "suggest roadmap changes", "improve my roadmap", "what is missing from my roadmap", "review my roadmap", or "should this be an initiative or ongoing work". Requires One Horizon MCP.

No reviews yet
0 installs
6 views
0.0% view→install

Install

$ agentstack add skill-onehorizonai-skills-one-roadmap-suggestions

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of One Roadmap Suggestions? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Roadmap Suggestions

Review a One Horizon workspace and suggest practical improvements to roadmap structure, sequencing, and scope.

Core rules

  • Default to suggestions only. Do not mutate roadmap items unless the user explicitly asks you to apply changes.
  • Ground every recommendation in current workspace data, not generic startup advice.
  • Prefer 3-5 well-supported suggestions over an exhaustive dump unless the user asks for a full audit.
  • If the roadmap is sparse, use planned work, blockers, bugs, and taxonomy to propose a starter roadmap.
  • Do not invent missing initiatives from weak signals. If the evidence is thin, label it as a hypothesis.
  • Do not recommend a work-type conversion unless you can explain why the current type is wrong and which type fits better.

Gather context

  1. Call list-initiatives with active statuses and includeHierarchy: true.
  2. Call list-planned-work with includeInitiatives: true.
  3. Call list-bugs for active bugs when defect pressure may imply missing roadmap investment.
  4. Call list-blockers with includeInitiatives: true.
  5. Call list-taxonomy when the workspace uses goals, products, companies, or components, or when category-like initiatives may really be taxonomy.
  6. If a recommendation depends on a specific item, call get-task-details before suggesting a reframe.
  7. Use search-tasks when you need to confirm a repeated theme across the workspace.

Decision rubric

Before giving a recommendation, classify it into exactly one of these buckets:

  • Hierarchy: parent-child reshaping, merge, dedupe, or split.
  • Work type: initiative vs taxonomy vs ongoing work vs bug vs feature request vs Todo.
  • Coverage: missing roadmap investment or missing foundational work.
  • Sequencing: roadmap order is likely wrong or incomplete.
  • Tagging: taxonomy is missing or misused.

Only surface a recommendation when at least one of these is true:

  • There are at least 2 supporting signals across initiatives, planned work, blockers, bugs, or taxonomy.
  • There is 1 strong signal from get-task-details that clearly contradicts the current roadmap shape.
  • The same gap or mismatch appears in multiple related tasks.

Heuristics

Hierarchy

  • Suggest parent and child reshaping when there are flat lists of tightly related initiatives, mismatched children, or child items that are really standalone efforts.
  • Suggest merging or deduplicating initiatives with overlapping titles, scope, or outcomes.
  • Suggest splitting initiatives that bundle multiple products, goals, or delivery tracks.

Work type

  • Suggest taxonomy when the item behaves like a reusable label or grouping concept such as a product area, customer, market, or goal rather than a deliverable.
  • Suggest ongoing work when the item is recurring and owner-driven with no clear end state.
  • Suggest a bug or feature request when the item is really issue intake rather than planned roadmap work.
  • Suggest a Todo only for small private follow-up work, not roadmap items.

Coverage and maturity

  • Suggest missing initiatives when blockers, bugs, or planned work cluster around the same gap.
  • Suggest sequencing changes when later work depends on missing foundations such as auth, testing, onboarding, or integrations.
  • Call out where the roadmap is under-tagged or under-grouped relative to existing taxonomy.

Response shape

  • Summary: 2-4 sentences on the current roadmap quality, maturity, and biggest pattern.
  • Suggestions: for each recommendation, use this format:
  • Type: one bucket from the decision rubric
  • Change: the exact change to make
  • Evidence: the concrete tasks, blockers, bugs, taxonomy, or hierarchy pattern that support it
  • Why: why this improves the roadmap
  • Confidence: High, Medium, or Low
  • Missing initiatives: include only when there are clear gaps, with concrete initiative titles.
  • Hypotheses: include only low-confidence ideas that may be useful but are not well supported yet.
  • Next step: ask whether to apply any subset using One Horizon tools.

Quality bar

  • Be direct. Do not pad the answer with roadmap theory.
  • Prefer specific initiative titles over abstract categories.
  • If the workspace already looks coherent, say so and suggest only the smallest useful changes.
  • If there is not enough evidence for a confident recommendation, say what is missing.

If the user wants changes applied

  • Use update-initiative for hierarchy, title, status, team, assignee, or taxonomy changes.
  • Use create-initiative for new roadmap items.
  • Use create-todo only when the recommendation is really a private follow-up.
  • Add a task comment when you materially reframe an existing initiative so the reasoning is visible.
  • Never rewrite descriptions just to log progress.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.