agentleFS
flowleap-aiflowleap-ai/flowleap-cli/skills/flowleap-tools/SKILL.md

flowleap-tools

Skill1 starsChanged 4 months agoMITView on GitHub

Discover and run FlowLeap backend tools through the agent-first /v1/tools facade — one stable contract (search_patents, get_bibliography, get_claims, get_patent_summary, compare_patents, …) instead of provider-specific endpoints. Trigger when an agent needs patent data and wants a uniform tool interface, needs to discover what the backend can do, or wants compound operations like patent summaries, comparisons, or prosecution timelines.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Search: search_patents (provider=epo_ops CQL | uspto Lucene), search_register (EP Register CQL, EP applications only — the one search with …

    line 83
  2. Very specific:

    Search queries are written by you — flowleap-patent carries the CQL method (term extraction, discriminating term, mandatory count probe).

    line 152
  3. Very specific:

    Never probe reachability with a search: it costs a provider call, needs a subscription and a patent-data key, and tells you nothing health …

    line 169

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

FlowLeap Tools (agent-first facade)

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. The connector and the app list the same registry under the same names; in the app a paid tool carries the mcp__flowleap__ prefix.

The Tools facade is the single agent surface for patent data: named tools invoked through /v1/tools, one success envelope, one error contract, a self-describing registry. It is not one API among several — every JSON patent-data capability is a tool.

flowleap tools is the direct way in. Every data command (patent, ops, uspto, citation, legal, npl, academic, analytics, ocr, patstat, and the one-call verbs) already runs on it, so use the named commands for ordinary work and reach for tools run when you want a parameter the command does not expose, a tool with no command of its own, or a schema you can read at runtime.

The per-source provider routes those commands used to call — the endpoints from before the facade became canonical — are retired endpoints: permanently removed, answering 410 Gone with a machine-readable successor. Retirement is forever, and a retired path is never reused. Only auth/OAuth, key validation and api request run outside the facade.

Discover

flowleap --json tools list                     # all tools + descriptions
flowleap --json tools describe get_bibliography  # JSON input schema for one tool
flowleap --json tools openapi                  # full OpenAPI document

Run

Inputs are JSON objects. Three equivalent styles — key=value pairs are easiest for flat inputs; values parse as JSON when possible (numbers, booleans, arrays):

flowleap --json tools run get_bibliography patent_number=EP1000000
flowleap --json tools run search_patents --input '{"query":"ti=\"solid state battery\"","range":"1-10"}'
flowleap --json tools run compare_patents --input '{"patent_numbers":["EP1000000","US10123456"]}'

A run answers one envelope, { success, tool, data, executionTimeMs, cached? }; the CLI prints data.

Errors — branch on the code

Failures answer { success: false, error: { code, message }, status }. Branch on error.code, never on message. Codes come from a closed registry and never change once shipped; message wording is freely editable by backend policy, so text matching is not a contract.

CodeStatusMeaning
INVALID_INPUT422Input failed schema validation; carries issues[]
UNKNOWN_TOOL404No tool by that name — re-run tools list
TOOL_EXECUTION_ERROR422The tool ran and failed
NOT_FOUND404The upstream provider has no such document
RATE_LIMITED / rate_limit_exceeded429Back off; carries retryAfterSeconds
INTERNAL_ERROR500Unexpected tool failure
subscription_required402A human must subscribe; carries upgradeUrl
data_keys_required / patent_provider_key_invalid400Key gate, each carrying provider. While stored keys are enabled, data_keys_required (and patent_provider_key_invalid when the rejected key is the stored key) also carries keysPageUrl and a human nextStep (store-epo-keys / store-uspto-key): give the human that FlowLeap "Patent-data keys" page link, and never ask for the key value in the chat — see flowleap-keys
odp_api_key_missing503USPTO ODP key gate
endpoint_gone410Retired endpoint; the build is stale — see below

A 410 endpoint_gone means this CLI build called a retired endpoint. The CLI exits 8 and prints an endpointGoneHint relaying the backend's successor and reason. Never retry the same call: retirement is permanent. Run flowleap upgrade, then flowleap skills update. Every request carries the Client version header (X-FlowLeap-Client: cli/<version>), which is observational only — logged as stale-client evidence, never used to reject a request.

Tool inventory

flowleap --json tools list is authoritative and discoverable at runtime — read it rather than trusting this snapshot when the two disagree.

Search: search_patents (provider=epo_ops CQL | uspto Lucene), search_register (EP Register CQL, EP applications only — the one search with a representative index: re= representative, op= opponent, nm= name in any role, surname-first, for when a role index returns 0; search_patents has none at any office), get_search_syntax (provider epo_ops | uspto | epo_register), search_uspto_portfolio_by_customer_number.

Retrieval, worldwide bibliographic (any publication number, EPO OPS): get_bibliography, get_abstract, get_family (INPADOC extended family), get_patent_family (simple family), get_legal_status, get_register_events, get_register_biblio (EP representatives and opponents), get_citations, get_patent_image (through MCP, include_images: true returns each PNG or JPEG page as an image content block you can see, with the page metadata in the text block), convert_patent_number.

Text, by office: get_claims (EP and WO from EPO OPS; US from BigQuery, in English only), get_description (EP and WO from EPO OPS), get_fulltext (EP and WO). Every other office answers FULLTEXT_UNAVAILABLE. For an office with no text source, use the family-member workaround in flowleap-shared.

US lookups and prosecution: get_us_grant, get_us_application, get_continuity, get_transactions, get_assignments, get_foreign_priority, get_patent_term_adjustment, get_attorney, get_application_documents, read_application_document.

Citations: search_office_action_citations (by application), search_enriched_citations (forward, by cited document), get_citation_stats.

Compound (one call, multiple sources): get_patent_summary, compare_patents (2-10 patents), get_prosecution_timeline, get_patent_term.

Literature: search_academic (Semantic Scholar + arXiv), search_npl (OpenAlex), get_npl_work (one cited paper to its OpenAlex record). Patent law is not a tool: read it from the office's own page with flowleap-legal.

Analytics and documents: patent_analytics (Topic Analytics, free-text keywords), ocr. Meta: server_info.

PATSTAT tools (the PATSTAT snapshot: no patent-data key, a gate and a rate limit per tool, data_edition and attribution on every result), each beside its command:

  • Portfolio and guarded SQL (flowleap-patstat): patstat_portfolio (patstat portfolio), patstat_query (patstat query), patstat_docs (patstat docs). patstat_query is sent exactly once on every path, tools run included; the agent owns its one retry (see flowleap-patstat).
  • Graph Analytics (flowleap-patstat-graph): patstat_resolve, patstat_cpc, patstat_patent, patstat_applicant, patstat_technology, patstat_neighborhood, patstat_path, patstat_explain (patstat graph resolve|cpc|patent|applicant|technology|neighborhood|path|explain).

flowleap-patstat carries the routing rule between Topic, Portfolio and Graph Analytics.

Recipes

Patent snapshot in one call:

flowleap --json tools run get_patent_summary patent_number=EP1000000

Portfolio triage:

flowleap --json tools run search_uspto_portfolio_by_customer_number customer_number=23456 limit=50

Search queries are written by you — flowleap-patent carries the CQL method (term extraction, discriminating term, mandatory count probe). The probe is the same tool with details=false, which skips the per-document bibliography fan-out — bare document references and the total you came for, at a fraction of the cost:

flowleap --json tools run search_patents query='ta="battery separator" AND ta="solid state"' range=1-1 details=false
flowleap --json tools run search_patents query='ta="battery separator" AND ta="solid state"' range=1-20

Reachability check — is the backend up, and which build?

flowleap --json health api    # public readiness probe; reports apiVersion

Never probe reachability with a search: it costs a provider call, needs a subscription and a patent-data key, and tells you nothing health api does not.

Auth

Auth, subscription, and rate limits: see flowleap-shared.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.