gleanwork

Search the public Glean developer documentation for API, SDK, MCP, authentication, indexing, and integration details. Use when the user is building with Glean rather than searching internal company knowledge — trigger phrases include Glean API, Glean SDK, integrate with Glean, indexing API, Glean Python or JavaScript SDK, and how to authenticate with Glean. Don't use it for internal company knowledge, which the using-glean skill covers.

Catch up on what you missed while away — meetings, decisions, action items, mentions, and important threads from a defined time window. Trigger phrases include "what did I miss", "catch me up", "I was out", "I'm back from PTO", "summarize while I was away", "what happened last week while I was off", "I just got back", "fill me in on what I missed".

Identify who owns, maintains, or has expertise in a specific code area — use when asked who to talk to about a system, who to request a code review from, or who has been actively working in a codebase area.

Gather architectural context about an internal system from code and documentation across the organization — use when asked to understand a system's architecture, find its repos, or get an overview before working on it.

Help the user connect a Glean MCP server so the Glean skills have tools to call — use when setting up Glean for the first time, adding another Glean server, checking whether Glean is connected, or troubleshooting missing Glean tools. Trigger phrases include "set up Glean", "connect Glean", "configure Glean MCP", "add a Glean server", "is Glean connected", "check Glean status", "Glean isn't working", "no Glean tools", "the Glean tools aren't showing up", "I installed the Glean plugin but nothing happens". Don't use it for actual enterprise queries once connected (use the using-glean skill); this skill only sets up and verifies the connection.

Find usage examples of an API, library, or code pattern across internal repositories — use when looking for how other teams use an internal API, library, or pattern before implementing something new.

Find subject matter experts on a topic, technology, or internal system by combining real expertise signals (code contributions, document authorship, and meeting or discussion activity) rather than relying on job titles. Use this whenever the user wants to know who to ask about something — phrasings such as who knows about, who is the expert on, the go-to person for, who has worked on, who can I ask about, or subject matter expert for. Use it even when the user does not say the word expert but clearly needs a person with deep, demonstrated knowledge of an area. Do not use it for plain directory lookups (an email address, manager, or org placement with no expertise angle), which is a simple employee lookup.

Prepare for an upcoming meeting — surface context from past meetings, related docs, and attendee information. Trigger phrases include "prep for my meeting", "context for the meeting", "what should I know before", "1:1 prep with", "prepare for [name]", "background on this meeting", "agenda context", "talking points for".

Synthesize the user's own work activity, priorities, and recent context using Glean — use when the user asks about their recent work, what they accomplished, what is urgent, what needs their attention, or wants help with status updates, 1:1 prep, or weekly summaries. Trigger phrases include what have I been working on, what did I do last week, what should I focus on, summarize my week, help me with my status update, what's blocking me, morning briefing, what happened while I was out. Do not use for general enterprise queries (use the using-glean skill), for someone else's activity, or for project-status questions where the user is not the subject (use a project-awareness skill).

Use Glean MCP tools to answer questions about company documents, internal wikis, policies, RFCs, design docs, people, teams, meetings, decisions, action items, email, calendar events, internal code, and the user's own work activity. Reach for this whenever the answer lives in enterprise systems rather than the local codebase or public web. Trigger phrases include "find the doc about", "what's our policy on", "where is the spec for", "who works on", "who owns", "find someone who knows", "what was decided in the meeting", "action items from", "search my email", "what did I work on last week", and "find the implementation in our repos". Don't use it for questions answerable from the local working directory, the public web, or generic programming knowledge.

Help the user connect a Glean MCP server so the Glean skills have tools to call — use when setting up Glean for the first time, adding another Glean server, checking whether Glean is connected, or troubleshooting missing Glean tools. Trigger phrases include "set up Glean", "connect Glean", "configure Glean MCP", "add a Glean server", "is Glean connected", "check Glean status", "Glean isn't working", "no Glean tools", "the Glean tools aren't showing up", "I installed the Glean plugin but nothing happens". Don't use it for actual enterprise queries once connected (use the using-glean skill); this skill only sets up and verifies the connection.

Generate a comprehensive handoff document so someone new can take over a project — current state, key people, essential knowledge, open items, and a first-30-days plan. Use when the user is handing off a project (trigger phrases include "generate a handoff", "hand off this project", "create a handoff doc", "I'm going on PTO / leaving / changing teams and need to document", "help someone take over") rather than a quick status read (use project-awareness for that).

Guide the user through authoring a new agent skill — when a skill is worth creating, the SKILL.md format, progressive disclosure, and common pitfalls. Also covers discovering skill opportunities from the user's work patterns and generating a SKILL.md from a description. Trigger phrases include "create a skill", "write a skill", "make a skill", "new skill", "SKILL.md format", "how do skills work", "skill best practices", "convert this workflow into a skill", "what should I automate", and "find skill opportunities". Don't use it for invoking existing skills — only when authoring a new one or understanding the skill format.

Discover and vet reusable skill opportunities from the user's Glean work patterns. Use when the user asks "what should I automate", "find skill opportunities", or wants to turn recurring work into skills. For authoring a known skill in Codex, hand off to the built-in skill-creator.

Run AI-driven tool-metadata optimization experiments for MCP servers using runVariantExperiment. Use when asked to improve tool descriptions, input schemas, or parameter descriptions, fix tool triggering, optimize tool discoverability, or run variant experiments against an eval dataset. Produces a measured, regression-guarded improvement proposal — never edits server source unless explicitly asked.

Generate LLM host simulation evals for MCP servers. Use when asked to test tool discoverability, write mcp_host evals, or validate tool descriptions with real LLM calls. Produces eval datasets and test runners where an LLM discovers and calls tools from natural language scenarios.

Reference guide for @gleanwork/mcp-server-tester — the Playwright-based testing and evaluation framework for MCP servers. Covers import paths, all 11 matchers, transport config, eval datasets, reporter setup, CLI commands, auth patterns, and common anti-patterns. Use when working with MCP server tests or evals.

Generate data-driven eval datasets for MCP server testing. Use when asked to create evals, evaluation datasets, or data-driven tests for MCP tools. Produces JSON eval datasets and Playwright test runners for @gleanwork/mcp-server-tester.

Generate Playwright tests for MCP server tools. Use when asked to write MCP tests, test MCP tools, or create test files for an MCP server. Produces direct-mode tests using @gleanwork/mcp-server-tester fixtures.

Use when extending pluginpack itself with support for a new AI app or plugin format — i.e. adding a target adapter to the pluginpack codebase. This is for contributors hacking ON pluginpack, not for end users packaging their plugins.

Use when someone wants to ship one set of agent skills/plugins to multiple AI apps (Claude Code, Cursor, Antigravity CLI, GitHub Copilot) from a single source, and needs to create or update a pluginpack.config.ts — choosing targets, source layout, MCP servers, and output directories.

Use when someone has an existing native plugin repo (a Claude Code marketplace, a Cursor plugin, an Antigravity CLI plugin, or a GitHub Copilot plugin) and wants to manage it from one portable source via pluginpack — generating the source layout and config, then proving the output matches the original with diff.

Use when working in a repo that uses pluginpack (it has a pluginpack.config.ts) and you need to build the native plugin outputs, validate them, prune stale files, or check whether a generated plugin repo has drifted — i.e. driving the pluginpack CLI and interpreting its exit codes.

How to create custom tools for the Glean Agent Toolkit. Use when building new tool integrations, extending the toolkit with custom Glean API calls, or creating new @tool_spec decorated functions. Triggers on: 'create tool', 'new tool', 'custom tool', '@tool_spec', 'tool_spec decorator', 'add a tool'.

How to use the Glean Agent Toolkit SDK. Use when building agents that integrate Glean enterprise search via the glean-agent-toolkit Python package. Triggers on: 'glean-agent-toolkit', 'glean agent toolkit', 'GleanContext', 'get_tools', 'as_openai_tool', 'as_langchain_tool', 'as_crewai_tool', 'as_adk_tool', '@tool_spec'.

Glean CLI: access company knowledge, search documents, chat with Glean Assistant, look up people, and manage enterprise content. Use when the user asks about internal docs, company information, people, policies, or enterprise data.

Discover and run Glean skills for enterprise app tasks

Use the langchain-glean package to call Glean from LangChain — chat models (ChatGlean / ChatGleanAgent), retrievers over Glean search and the people directory, and Glean tools for LangChain agents. Load when writing Python that imports langchain_glean or wires Glean into a LangChain app, chain, or agent.

Triage and fix broken-link reports from the Glean developer site's nightly link checker. Use when investigating a `Broken Links Detected` issue, a failing `Link Checker` workflow run, or any scenario where you need to act on a `link-check-fixme.md` / `link-check-results.json` artifact. Covers source-content fixes, adding exclusions, and verification.

Use the glean-mdm CLI to provision Glean's AI-coding-tool integration on managed devices (IT/MDM fleet tooling) — install the Glean editor extension and configure MCP servers across supported AI coding tools for every user on a machine, on a system schedule. Load when operating, scripting, or troubleshooting glean-mdm on a managed device.