Use when writing up a data analysis: conclusion first, then evidence, with definitions, method, limitations and recommendations.
w01fb0ss/PiBot
Open-source, local-first AI agent team for your desktop: multi-agent group chats, computer use, browser, MCP, routines and approvals. macOS, Windows, Linux + Android. Built on pi.
- 37 个技能
- 116 Star
装得最多的 5 个
Use when adding or changing an HTTP / RPC API: check naming, validation, errors, pagination, compatibility and security.
Use to prepare data for a periodic brief: align definitions, summarize tables, and record items left out.
Use after finding or reproducing a problem: write a bug report anyone can reproduce from.
Use when fixing a bug: reproduce it, find the root cause, make the smallest fix, then verify and report.
其余 32 个技能
Use when delivering a change: state what changed, how it was verified and what is left.
Use when making or checking charts, formatting Excel reports, filling templates or preserving existing workbooks; check definitions, formula caches and native objects.
Use when reviewing code changes: check correctness, security, maintainability, performance and tests, and label findings by severity.
Use when comparing competitors or alternatives: pick dimensions, gather sourced facts, build a comparison table, conclude with opportunities.
Use when preparing topics, platform-specific copy, visual requirements and a publishing checklist around one theme.
Use for marketing copy, product descriptions, ads and headlines: settle audience and goal, then write 2–3 angles with a framework and several headlines.
Use when building classroom courseware to sequence answers, annotations, pointers and animations in teaching order.
Use on a new CSV / spreadsheet: run the bundled script to see rows, columns, missing values, duplicates and value ranges before analysing.
Use to review narrative, layout, contrast, spacing, typography, charts and notes, with a slide-by-slide improvement list.
Use to turn content actually read from Word, PDF or web sources into a sourced, editable presentation.
Use to inspect an existing PPT, list improvements, edit individual objects and recheck while preserving unrelated content.
Render and check every slide before delivery; collect notes in the built-in Bot browser and revise only the annotated slides.
Use to define the purpose, audience and duration and create an assertion-led storyline with a specification for every slide.
Use for implementation: scale scope, delivery steps and verification to the size of the change.
Use when writing an email: clear purpose and recipient, one topic per email, draft for the user to approve; sending needs approval.
Verify a claim against at least three independent sources; report a verdict, confidence and traceable provenance.
Read and send Gmail / 读取和发送 Gmail 邮件
Use when writing a scene description (prompt) for an image model: subject, setting, composition, light, style, colour and ratio, with a few directions.
Use when turning meeting notes, transcripts or chat into minutes: decisions, action items with owner and due date, open questions.
Use for a weekly or monthly brief written from one folder of material and compared with last period.
Use when preparing content for a social or publishing platform: adapt titles, length, layout, hashtags, visuals and timing to its habits.
Build and locally test a Pi Bot connection plugin with the bundled SDK.
Review plugin safety and installation readiness.
Use when editing, tightening, shortening or rewriting text: keep facts and meaning, adjust to the goal, and say what changed.
Use when researching a topic and writing it up: split the question, find sources, separate fact from inference, cite every conclusion with confidence.
Use when delivering a review verdict: verdict, required fixes and suggestions in a report file.
Use when code handles user input, permissions, secrets, file paths, command execution or outbound requests: check for common security problems.
Use when reading articles, papers or reports: keep one card per source (citation, key points, short quotes, credibility) for later citing.
Use when designing test cases for a feature or change: cover happy path, boundaries, bad input, permissions, concurrency and regression.
Use when reporting test results: one row per case with expected, actual and outcome, then a delivery verdict.
Use when recording to-dos, tracking due dates or setting up recurring reminders: write the to-do, propose a routine when needed, and check connected calendar or contacts.
Use for a personal or project weekly report: distil progress, issues and next week's plan from the material, results first.