菜单
Send messages to people via chat platforms (e.g. Microsoft Teams, Slack) using browser automation. Resolves people by name using the knowledge graph, drafts messages for approval, and sends via the web app. Use when the user asks to message, ping, or chat with someone.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Follow a live Anarlog session in real-time, coaching the user through a meeting or interview. Understands context from the session title, knowledge base, and candidate pipeline. Provides talking points, flags gaps in coverage, and suggests follow-up questions as the conversation unfolds. Use when the user asks to follow, shadow, or coach them through a live meeting.
Process Anarlog meeting sessions (memos, summaries, transcripts) into the knowledge graph. Extracts people, organizations, projects, and topics from AI-generated meeting summaries and user notes, creating or updating Obsidian-compatible notes in Knowledge/. Use when the user asks to process meeting notes or after Anarlog sessions.
Trim a Anarlog transcript to its logical end. Recordings are often left running after a meeting finishes — this skill finds the natural conclusion (goodbyes, sign-offs) and cuts the transcript there. Use when the user asks to trim, cut, or clean up a Anarlog transcript.
Generate an A4 single-page HTML candidate assessment report benchmarked against the agent-aligned engineering standard. Use when the user asks you to create a candidate report, one-pager, or visual assessment for a hiring manager.
Record the knowledge-graph changes made during the current session into a single shared Knowledge/CHANGELOG.md so the team can see what changed and why. Use when the user asks to log, record, or write up the changes they just made to the knowledge base — typically at the end of a session of edits.
Generate PDF slide decks from user requests using Playwright to render HTML slides to PDF. Use when the user asks to create a presentation, slide deck, or pitch deck. Pulls context from the knowledge base for company info, project details, and people.
| name | send-chat |
| description | Send messages to people via chat platforms (e.g. Microsoft Teams, Slack) using browser automation. Resolves people by name using the knowledge graph, drafts messages for approval, and sends via the web app. Use when the user asks to message, ping, or chat with someone. |
| compatibility | {"requires":["browser-automation"]} |
| license | Apache-2.0 |
| metadata | {"version":"3.12.0","author":"forwardimpact"} |
Send chat messages to people using browser automation against a web-based chat platform (Microsoft Teams, Slack, or similar). Resolves recipients by name from the knowledge graph so the user can say "message Sarah about the standup" without needing exact display names.
Run when the user asks to:
BEFORE messaging anyone, you MUST look up the person in the knowledge base.
When the user mentions ANY person:
rg -l "{name}" Knowledge/People/This context is essential for:
The user will refer to people by first name, last name, or nickname. Resolve to a full name using the knowledge graph:
# Find person by partial name
rg -l -i "{name}" Knowledge/People/
# If ambiguous, read candidates to disambiguate
cat "Knowledge/People/{Candidate}.md"
If ambiguous (multiple matches), ask the user which person they mean — list the matches with roles/orgs to help them pick.
If no match, tell the user you don't have this person in the knowledge base and ask for their full name as it appears in the chat platform.
Every message MUST be drafted as a text file first. This ensures the user can review and edit the exact message before it's sent.
Drafts/chat-{recipient-slug}-{date}.md
{recipient-slug} = lowercase, hyphenated full name (e.g. sarah-chen){date} = ISO date (e.g. 2026-02-19)Draft file format:
To: {Full Name}
Via: {Platform name}
Date: {YYYY-MM-DD}
---
{message body}
The message body (everything below the --- separator) is what gets pasted into
the chat.
Message guidelines:
•, —) to keep structure. Multi-line
formatting is unreliable via browser automation.Once the user has approved the draft, send it as a single submission — paste the entire message at once rather than typing line by line.
Check which platform is available:
If the person doesn't appear in search, inform the user — they may not be in the same organization.
---)If the message is substantive (not just "hey" or "thanks"), note the interaction on the person's knowledge note:
- {YYYY-MM-DD}: Messaged on {Platform} re: {topic}