| name | find-template |
| description | Recommend the best-fit workspace template or plugin variant for a described use case, project idea, or research question. Use when the user asks "what template should I use for X", "is there a scaffold for Y", "which workspace fits Z best", "recommend a template for…", or describes a project without naming a template. Returns a short ranked shortlist with rationale — does not create the repo. If the user wants to proceed, hand off to the matching cluster plugin's `new-workspace` skill. |
Find Template
Propose the best-fit workspace scaffold for a described use case. This is an advisory skill — it does not create a repo. When the user then wants to spin one up, delegate to the appropriate cluster plugin's new-workspace skill (e.g. /research-space:new-workspace, /workspace-foundational:new-workspace, etc.) with the chosen variant.
Background
This skill absorbs the recommender from the retired new-repo-from-template plugin. In the new model, each of the ~27 cluster plugins in the danielrosehill marketplace bundles its own variants; this skill's job is to route the user to the right plugin + variant, not to enumerate 111 standalone template repos.
Inputs
The user describes what they want to do — e.g.:
- "what template should I use to research distribution platforms for Suno-created music?"
- "got a template for tracking expenses across multiple family members?"
- "which scaffold fits best for a single-company deep dive on a potential acquisition target?"
- "I want to map the AI coding assistant ecosystem — anything for that?"
- "I need to parse a stack of PDF reports and pull out the key claims."
Extract the intent (what the workspace is for), not just surface keywords.
Steps
1. Classify the intent
Bucket the request into one of:
- Research / investigation → likely
research-space plugin
- Report parsing / document triage →
workspace-foundational (report-parsing variant)
- Context / RAG maintenance →
workspace-foundational (context-management variant)
- Inventory / asset tracking →
workspace-foundational (inventory-analysis variant)
- Web analytics →
workspace-foundational (web-analytics variant)
- Career / job search / salary →
career plugin
- Purchasing / procurement — a decision between known options →
procurement-tools plugin (general-purchasing variant)
- Procurement research — "what's on the market that meets this spec?" →
procurement-tools plugin (market-landscape variant)
- "I don't know what this thing is called or what to search for" →
procurement-tools plugin, define-spec command (spec + annotated search vocabulary, accepts a photograph)
- Region-specific shopping →
shopping plugin
- Knowledge base / wiki / SOP →
knowledge-documentation plugin
- Content writing / blog / thinkpiece →
content-writing plugin
- Household / personal finance →
personal-finance plugin
- Filesystem organisation →
filesystem-organiser plugin
- Brainstorming / idea generation on a subject →
ideation-planning plugin (ideation-session variant)
- Deep evaluation of one idea →
ideation-planning plugin (single-idea-eval variant), or the business-idea-eval plugin when the idea is commercial
- Ranking or shortlisting many candidates →
ideation-planning plugin (multi-idea-ranking variant)
- Low-friction idea inbox →
ideation-planning plugin (idea-capture variant)
- Choosing between defined options →
decision-evaluation-framework plugin
- Generic ongoing workspace →
workspace-foundational (generic-workspace variant)
If the request genuinely doesn't fit, say so and suggest the generic-workspace variant as a starting point.
1b. Disambiguate decision vs survey
Several clusters split along the same seam, and it's the commonest way this skill routes wrong. Ask which side the user is on whenever it isn't explicit:
- "Which of these should I get?" — the options are known, the output is a pick. →
purchasing:general-purchasing
- "What's even out there that does this?" — the spec is known, the options aren't, the output is a map of the field and its gaps. →
purchasing:market-landscape
Signals for the survey side: the user states a requirement rather than a product; asks what the "lay of the land" is; doesn't know whether the thing exists; wants to know what's available currently; or says they're not buying yet. A survey can legitimately conclude "this doesn't exist" — if that outcome would be useful to them, they want the landscape variant.
The same seam separates research-space:ecosystem (mapping a technology or vendor ecosystem for understanding) from purchasing:market-landscape (mapping a market against a buying spec, with conformance and price bands). If there's a spec with must-haves behind the question, it's the purchasing one.
2. Return a shortlist
Output 2–4 ranked recommendations. Each entry:
**<plugin-name>:<variant>** — <one-line purpose>
Why: <rationale tying the variant's purpose to the user's stated intent>
- Top pick with a 1–2 sentence rationale.
- Runner-ups with one line each on why they might also fit and why they're ranked below.
- If nothing fits well: suggest
workspace-foundational:generic-workspace and note the user could always adopt a more specific cluster later.
3. Offer the next step
End with a single line offering to proceed: "Want me to spin up <plugin-name>:<variant> as a new repo? (I'll hand off to /<plugin-name>:new-workspace.)"
Do not auto-invoke any new-workspace skill — wait for the user to confirm.
Guardrails
- Never invent plugins or variants that don't exist in the danielrosehill marketplace.
- If the user's description is ambiguous, ask one clarifying question before recommending (e.g. "Is this a one-off report or an ongoing workspace?").
- Don't dump the full cluster map — shortlist only.
- Prefer narrower variants (e.g.
inventory-analysis) over generic ones (generic-workspace) when the intent clearly fits.
