opencode-skills

Send Slack direct messages through the web UI using headed agent-browser, with the human logged in interactively. Supports a single person or batching the same message to many people one by one without creating a group chat. USE FOR: send a Slack DM, send a message to X on DM, message these people on Slack, DM this update to a list of people, send the same Slack message to multiple recipients, batch Slack DMs, team communication update, stakeholder update, notify people individually on Slack. DO NOT USE FOR: Slack API, Slack bots, webhooks, desktop Slack automation, channel announcements, group chats, emails, calendar invites.

Generates a personalized 1:1 prep/debrief by reading a person's vault profile, recent call notes, GitHub activity (PRs, commits, issues), and project board status. Produces an agenda with context, talking points, and follow-ups. Use when the user says 'prepare my 1:1', 'debrief for [name]', '1:1 prep', 'what should I discuss with [name]', 'catch me up on [name]', 'prep for my meeting with [name]', or mentions preparing for a direct report conversation. Do NOT use for delivering feedback (use feedback skill) or processing meeting transcripts (use read-meeting-notes).

Processes meeting transcripts from configured sources (Zoom, Teams, local files, calendar recordings). Extracts action items, decisions, and key topics, then creates structured call notes in the vault routed to the relevant person or meeting folder. Use when the user says 'process meeting notes', 'read my transcripts', 'import meeting recording', 'download zoom transcript', 'process today's meetings', 'what happened in my meetings', 'create notes from transcript', or mentions meeting recordings or transcripts. Do NOT use for 1:1 prep (use debrief) or manual note-taking.

Search the web using a local SearXNG metasearch engine via MCP. Use this skill when the user asks to search the web, look something up online, find information on the internet, research a topic, check recent news, or when you need up-to-date information that is not in your training data. USE FOR: web search, internet lookup, finding recent information, researching topics, checking news, discovering URLs, finding documentation online, verifying facts against current sources. DO NOT USE FOR: searching the local codebase (use Grep/Glob instead), reading known URLs (use WebFetch instead), or querying internal databases.

Upload Markdown files as Google Docs. Use when the user asks to create, upload, publish, or push a markdown file to Google Docs/Drive. Supports creating new docs and updating existing ones. Converts Markdown to HTML then uploads with Google Drive mimeType conversion to native Google Docs. Authenticates via gcloud CLI. Supports both user credentials and ADC.

Crawl and extract content from web pages using Crawl4AI via MCP. Use when the user asks to crawl a website, extract content from a URL, scrape a page, get a screenshot of a site, generate a PDF from a URL, convert a webpage to markdown, extract structured data from a page, or run JavaScript on a remote page. USE FOR: web crawling, page scraping, content extraction, webpage screenshots, PDF generation, JavaScript execution on pages, converting URLs to markdown, deep crawling multiple pages. DO NOT USE FOR: web search (use searxng-search), reading a known URL when simple fetch suffices (use WebFetch), local file operations.

Create and export Excalidraw diagrams — architecture diagrams, communication diagrams, flowcharts, system overviews. Use this skill whenever the user asks to create, update, modify, or export any kind of visual diagram, drawing, or architecture visualization. Also use when the user mentions boxes, arrows, connections between components, or wants to visualize system architecture, data flows, or service interactions. Handles both creating .excalidraw JSON files programmatically and exporting them to SVG/PNG server-side without a browser.

Convert Google Docs files to Markdown. Use when the user asks to convert, export, or read a Google Doc as markdown. Also fetches open comments from a Google Doc (comments command). Supports Google Docs URLs, document IDs, and searching for documents by name on Google Drive. Authenticates via gcloud CLI. Supports both user credentials and ADC.

Compact the current conversation into a handoff document for another agent session to pick up. Use when the user says "handoff", "hand off", "save context", "write a handoff", "summarize for next session", or wants to capture the current session state so a fresh agent can continue the work.

Convert PDF files to clean, structured Markdown. Use when the user wants to read, extract, analyze, or convert a PDF document to markdown. Triggers on requests like "convert this PDF", "read the PDF", "extract text from PDF", "load the PDF into context", or when the user mentions a .pdf file and wants its content as text or markdown. Handles tables, headers, lists, and images.

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.

Structured cross-repo or cross-system investigation with consistent deliverable format. Use when user asks to research, investigate, compare, or analyze something across multiple sources, repos, tools, or systems. Also use when user asks "how does X work across our repos", "what patterns do we use for Y", "compare approaches to Z", "investigate options for", "what are best practices for". Trigger phrases include "research", "investigate", "deep dive", "compare", "analyze across", "survey", "what are our options".

Rewrite AI-generated Polish text to read as naturally human-written, bypassing AI detectors (GPTZero, ZeroGPT, Originality.ai). Applies Polish-specific word replacements, sentence rhythm variation, structural de-patterning, register correction, and tonal adjustments targeting how detectors score Polish text. Use when user says "humanizuj", "humanizuj tekst", "popraw tekst AI po polsku", "make Polish text sound human", "humanize Polish", "za bardzo brzmi jak ChatGPT", "tekst brzmi sztucznie", "rewrite Polish AI text", "usuń ton AI", or pastes Polish text that reads like typical LLM output. Also use when user asks to edit Polish text for natural voice or authenticity in the context of AI-generated content.

Rewrite AI-generated text to read as naturally human-written, bypassing AI detectors (GPTZero, Originality.ai, Turnitin). Applies word-level replacements, sentence rhythm variation, structural de-patterning, and tonal adjustments grounded in how detection algorithms actually score text. Use when user says "humanize", "make this sound human", "rewrite to pass AI detection", "remove AI tone", "humanize AI text", "make less robotic", "AI detector", "sounds too much like ChatGPT", "rewrite naturally", or pastes text that reads like typical LLM output. Also use when user asks to edit text for "voice", "authenticity", or "natural tone" in the context of AI-generated content.

Explore a codebase to find opportunities for architectural improvement, focusing on making the codebase more testable by deepening shallow modules. Use when user wants to improve architecture, find refactoring opportunities, consolidate tightly-coupled modules, or make a codebase more AI-navigable.

Install, list, update, or remove OpenCode skill repositories. Use this skill whenever the user wants to add a new skill repo, install skills from a GitHub URL, add a skills repository, set up a new skill source, or mentions a GitHub repo containing SKILL.md files that they want available in OpenCode. Also use when the user asks to list, update, or remove installed skill repos. Trigger phrases: "add this skill repo", "install skills from", "add skill", "remove skill repo", "update skill repos", "list skill repos".

Search and retrieve OpenCode session history to find which session and directory was used for a specific task. Use this skill whenever the user mentions past sessions, asks where they worked on something, wants to find a session by PR URL, branch name, commit message, topic, repo name, or any keyword. Also trigger when the user wants to recall recent activity, trace back where a change was made, find which directory they were in, or look up what they did on a particular date. Even if the user doesn't say "session" explicitly -- if they're asking "where did I do X", "when did I work on Y", "find the PR I made for Z", or "what was I doing last week", this skill applies.

Detect complex or bulk tasks and nudge toward OpenSpec workflow. Use proactively when user asks to do something across multiple repos, migrate or refactor a large surface area, create many issues or PRs at once, do bulk operations on 5+ entities, implement a multi-step feature touching several files or systems, or when single-shotting has already produced corrections. Trigger phrases include "for each repo", "across all", "migrate", "bulk", "batch", "rollout", "for every", "update all repos".

Verify accuracy after bulk operations, aggregations, or multi-entity outputs. Use proactively after generating lists, tables, counts, summaries across multiple sources, completing bulk file edits, creating multiple issues or PRs, producing any output where count accuracy matters, or when the user asks to "verify", "double-check", "audit", or "make sure the numbers are right". Trigger on bulk output completion, aggregation results, multi-source summaries. Also use for loop detection, stuck detection, session spiral, retry limit, am I stuck, not making progress, tried this before. Auto-triggers when same error appears 3+ times, session exceeds 500 messages without deliverable, user sends correction signal 3+ times on same topic, agent spawns >10 subagents for single investigation, or any retry of a previously-failed approach.

Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, or asks for test-first development.