recipe-custom-dashboard
Skill1 starsChanged 4 months agoMITView on GitHub
Presentation-craft recipe for turning verified FlowLeap patent data into a disposable, self-contained HTML dashboard — the agent writes one small zero-dependency Node program per question that fetches through the CLI, computes every number in code, and emits an offline single-file dashboard with inline SVG charts and a full provenance footer. Trigger when the user asks to make, build, generate, or visualize a dashboard, chart, or visual report of patent data — a company's portfolio, filing trends across competitors, a technology landscape / white-space map, or citation impact.
In this repository34 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- flowleap-academicSkill
- flowleap-authSkill
- flowleap-citationSkill
- flowleap-keysSkill
- flowleap-legalSkill
- flowleap-nplSkill
- flowleap-opsSkill
- flowleap-patentSkill
- flowleap-patstat-graphSkill
- flowleap-patstatSkill
- flowleap-sharedSkill
- flowleapSkill
- flowleap-toolsSkill
- flowleap-usptoSkill
- persona-ip-analystSkill
- persona-patent-attorneySkill
- persona-researcherSkill
- persona-startup-founderSkill
- recipe-academic-literature-reviewSkill
- recipe-audit-reportSkill
- recipe-claim-analysisSkill
- recipe-claim-draftingSkill
- recipe-find-betterSkill
- recipe-freedom-to-operateSkill
- recipe-infringement-chartingSkill
- recipe-invalidity-analysisSkill
- recipe-invention-disclosureSkill
- recipe-maintenance-feesSkill
- recipe-office-action-responseSkill
- recipe-patent-landscapeSkill
- recipe-patent-to-reportSkill
- recipe-prior-art-searchSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 11
Instead of a fixed dashboard product, write a small program per question: fetch verified data through the FlowLeap CLI, compute every …
- Distinctive: line 30
Numbers only from code — including prose: Every value in the HTML (chart, table, and every narrative sentence) is interpolated from a …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Recipe: Custom Dashboard (Disposable, Verified-Data)
Instead of a fixed dashboard product, write a small program per question: fetch verified data through the FlowLeap CLI, compute every number in code, and emit one self-contained HTML file. This skill owns presentation craft only — the analysis (which query, which corpus, how to read the result) lives in the analytical recipes. Do the analysis there first, then render here.
In a chat client with the FlowLeap connector or in the FlowLeap Patent AI app, call the tools named in the flowleap-shared tool table instead of these commands, and never run a flowleap command there. Neither host runs generate.mjs, because it fetches through the CLI: describe the data, and hand the numbers to the user exactly as the tools returned them, with each call's tool, parameters, and dataset identity.
Vocabulary (agent-v2 CONTEXT.md): Portfolio Analytics = grounded PATSTAT
aggregates for one harmonized applicant (data_edition stamped). Topic
Analytics = full-corpus filing analytics over the quarterly Google-Patents slice
(corpus-slice stamped). Verified-Data Contract = every figure on the page is
interpolated from a computed variable, never typed by the model.
Mandatory rules
These are product decisions, not style preferences. A dashboard that breaks one of them is wrong even if it looks right.
- Numbers only from code — including prose. Every value in the HTML (chart,
table, and every narrative sentence) is interpolated from a computed variable.
The model never literal-types a figure. Use the template's
renderNarrative(computed)helper for sentences; embed backend-authoredsummarystrings verbatim as canonical captions. In chat, the dashboard is the source of record: the script prints a summary block — quote its numbers, never re-type them from memory. - Data via CLI subprocess only. Scripts spawn
flowleap … --jsonwithchild_process.execFile. Never hold a token, never raw-fetchthe backend. The binary is injectable viaFLOWLEAP_CLI_BIN(defaultflowleap) so the program stays testable, but it is always a subprocess. - Node-only, zero-dependency. One self-contained
.mjsper dashboard (Node ≥ 18). Nonpm install, no build step, no chart library. The inline SVG chart helper is copied into each program — keep it copy-in, not imported. Bundle layout. Workspace-relative
dashboards/<subject-slug>/:generate.mjs— the program. Refinements edit this file and re-run.dashboard.html— regenerated output. Never hand-edited.data/— one raw-JSON sidecar per backend call.
Slugs are subject-based, no timestamps (
dashboards/siemens-portfolio/, notdashboards/2026-07-23-run/). Re-running refines in place; to keep a prior state, copy the bundle or commit it first. Never overwrite a bundle you did not create without showing the user its current contents.- Provenance footer, per data source. For each backend call: the tool/verb +
its parameters, the dataset identity (
data_editionfor Portfolio Analytics; the corpus-slice stamp for Topic Analytics), and retrieved-at. A source with no dataset identity gets an explicit "dataset identity unavailable" row — never omit it. Plus one page-level generation timestamp and the CLI/backend versions. - Reproduce block. The HTML embeds the exact commands/script that produced the
data, including any pinned exact applicant name. Applicant ambiguity is
resolved once with the user and then pinned as a constant in
generate.mjs. Never auto-pick a 422 candidate — a non-interactive run exits non-zero listing the candidates, and the affected section is not rendered. - Self-contained HTML. Inline CSS + inline SVG only. No CDN, no external fonts,
no runtime JS dependency. It must open from
file://offline. - Local by default, never auto-published. Dashboards can contain pre-filing
invention data. Open in the system browser (
open/xdg-open). Publishing to any hosted surface happens only on explicit user request. In-IDE viewing:file://will not load in the IDE's Simple Browser — run a one-line static server (python3 -m http.server 8080from the bundle dir) and openhttp://localhost:8080/dashboard.html.
The four templates
Copy the matching template from references/ into dashboards/<slug>/generate.mjs,
pin the constants at the top, and run it. Each is a single runnable .mjs with the
data contract documented in its header comment and the SVG chart helper embedded.
| Template | Question | Data source | Analytics kind |
template-portfolio.mjs | One company's filings by year × office + grant ratio | flowleap patstat portfolio | Portfolio |
template-filing-trends.mjs | Several companies' filings over time, compared | flowleap patstat portfolio (per applicant) | Portfolio |
template-landscape-whitespace.mjs | CPC × year heatmap → white space in a technology area | flowleap analytics (scoped per CPC) | Topic |
template-citation-impact.mjs | Forward-citation distribution for a patent (how influential) | flowleap citation forward / stats | — |
Portfolio and filing-trends need the PATSTAT layer. Landscape and citation-impact do
not. Degrade gracefully: if patstat returns patstat_unavailable, the template
exits non-zero and tells the user that the landscape and citation-impact dashboards
remain usable — it does not fail the whole recipe.
Chart craft (baked into the helper)
The copied helper enforces the house style so you do not have to: an Okabe–Ito
colorblind-safe categorical palette, a single-hue sequential ramp for heatmaps, a
white background, readable axes with units and year labels, thousands-separated
numbers, and no gradients or chartjunk. Every chart carries its title and a
Source: label inside the SVG. Do not hand-tune colors per chart — extend the helper
if a new mark is needed.
Reproduce & provenance (what the footer must contain)
Generated: <ISO timestamp> # one per page
CLI / backend: flowleap <ver> / <backend ver> # flowleap --version, flowleap --json doctor
Per source:
verb + params e.g. patstat portfolio "Siemens AG" --from-year 2015 --to-year 2024
dataset identity data_edition: "PATSTAT Global 2025 Spring" (or corpus-slice stamp)
— or the literal "dataset identity unavailable"
retrieved-at <ISO timestamp>
Reproduce: the exact commands, with the PINNED applicant constant, copy-pasteable
Data: links to ./data/*.json sidecarsThe templates build every one of these fields from computed variables and the
response envelope — read renderFooter() in any template and keep its markers.
Routing
A standalone trigger ("make me a dashboard of X") routes here for the rendering,
but the analysis belongs to a matching recipe: portfolio/filing questions →
Portfolio Analytics data; landscape/white-space → recipe-patent-landscape (this is
its final rendering step); citation impact → the flowleap-citation family. Run the
analysis, pin the resolved parameters, then render with the template.
Output
- A
dashboards/<subject-slug>/bundle:generate.mjs,dashboard.html,data/. - An offline, single-file
dashboard.htmlwith inline SVG charts and the full provenance + reproduce footer. - A printed summary block (computed numbers + file paths) for the agent to paste in chat — the quoted numbers come from the script, never re-typed.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

