sensible-docs

Generates and publishes a monthly Sensible product changelog. Invoke whenever the user wants to write, draft, or publish a monthly changelog, or mentions writing up a release summary given a list of PRs. The user will provide PR numbers (from the frontend or backend GitHub repo, or both) and optionally a list of doc topic URLs. The skill structures the input, fetches PR context, drafts the changelog in Sensible's house style, gets user approval, then publishes as a hidden draft to readme.com.

Creates or updates OpenAPI JSON spec files in the /reference directory of sensible-docs, based on backend PRs from sensible-hq/sensible. Use this skill whenever the user provides PR numbers or GitHub URLs from the backend repo and wants to document new API endpoints, update existing endpoint schemas, or add new fields to an existing API reference. Also triggers when the user mentions "new API", "update the spec", "API reference from PRs", "write a spec for [feature]", or provides a list of PRs and asks to document them. This skill handles the end-to-end workflow: reading PRs, extracting API surface changes from diffs and Zod/TypeScript schemas, drafting OpenAPI 3.0.3 JSON, creating supporting markdown files, and updating _order.yaml navigation files.

Given a sensible-hq/sensible PR number, analyze the engine/API changes and update the sensible-docs repo accordingly, then open a PR. Handles both updating existing pages and creating new pages. If you already know which type of change is needed, use update-existing-doc or create-new-doc directly for a more focused workflow.

Generates a "how to extract data from X" blog post draft for Sensible. Takes a document type as input and either fetches the matching prebuilt config from the open-source library, or accepts an explicit config file/URL and optional PDF path via --config and --pdf flags. Invoke whenever the user wants to write or draft a "how to parse/extract X" blog post.

Checks Sensible documentation files for terminology violations before publishing. Invoke whenever the user wants to check, review, validate, or proofread a Sensible doc (changelog, reference page, integration guide, concept topic), or asks whether a file uses correct terminology. Accepts a file path, pasted content, or both. Also invoke if the user asks to "run the checker" or "check terminology".

Generates or refreshes the Sensible integration guide template by reading existing integration guide pages from the published docs and analyzing their conventions. Use this skill whenever asked to generate, update, or refresh the integration guide template, or when .claude/style-guide/integration-guide-template.md is missing or stale. Also invoke before writing a new integration guide from scratch.

Enriches JSON5 code blocks in a Sensible doc or blog post draft with canonical inline comments. Reads the comment reference, finds every fenced code block in the target file, and adds the canonical comment for each parameter that has one. Does not invent comments — only uses comments from the reference file. Use this skill before publishing any doc or blog post that contains SenseML code examples.

Walks a developer through the Sensible + QuickBooks Online integration from setup to a working bill. Invoke whenever the user wants to set up the QuickBooks integration, run or follow the QuickBooks tutorial, connect Sensible to QuickBooks Online, automate invoice extraction into QuickBooks, or create bills from vendor invoice PDFs using Python. Use this skill even if the user says they "already have credentials" or are "partway through" — it handles all starting points.

Generates or refreshes the Sensible SenseML docs style guide by reading existing reference pages from the published docs and analyzing their conventions. Use this skill whenever asked to generate, update, or refresh the style guide, or when the style guide files at .claude/style-guide/ are missing or stale. Also invoke before writing a batch of new SenseML reference pages from scratch.

Given an extractor directory path from the sensible engine repo, create or update a full-fledged documentation example — copying the golden PDF, generating a screenshot, assessing and redacting PII, and writing the example section with JSON5 comments.

Create a new concept topic page in docs/Senseml reference/concepts/ — for explanatory "how it works" topics, not reference pages. Use this skill when you need to document a concept, behavior, or feature overview (not a new method, preprocessor, or config option). If you're documenting a new method or preprocessor introduced by a PR, use create-new-doc instead.

Revise a SenseML reference doc draft to match the house style. Applies rules for intro sentences, metadata, examples structure, JSON5 comment placement, and parameter descriptions.

Given a sensible-hq/sensible PR number, create new sensible-docs reference pages for new methods, preprocessors, or other features introduced by the PR — then open a PR. Use this skill when you already know the PR adds something new that needs a new doc page. If you're not sure whether to create or update, use update-docs-from-pr instead.

Given a sensible-hq/sensible PR number, update existing sensible-docs pages to reflect changes to existing features, parameters, or behaviors — then open a PR. Use this skill when you already know the PR only affects existing docs (no new pages needed). If you're not sure whether to create or update, use update-docs-from-pr instead.

Given a sensible-hq/sensible PR number, find any new/modified extractor test directories and upload their configs and golden PDFs to the Sensible account via the API.