heading-os

Daily email intelligence processor. Scans incoming + outgoing 31C Exchange emails (ceo@31c.io via exchangelib/EWS), groups by conversation thread, categorizes for CRM actions, tasks, pipeline updates, knowledge capture, and relationship signals. Presents digest for approval before executing. Integrates with /prime for morning briefing. Trigger: '/email-intel', 'process emails', 'email digest', 'check my email'. Do NOT trigger for: sending email (use send-email.py), reading specific emails, email search queries, or Gmail operations.

Ultrathink principal-engineer review gate. Runs a Validate-Identify-Improve-Adjust (VIIA) pass over a target - a plan awaiting approval, just-executed work, a specific file or directory, or the entire workspace - then runs an adversarial refutation layer (Phase 2.5) with cross-family judge rotation (Claude / Gemini / Grok) and optional two-agent debate on BLOCKER + HIGH findings before presenting findings with concrete proposed fixes for batched approval. Blocks forward progress until approved. Triggers on "scrutinize", "stress-test this", "principal review", "validate and improve", "review the plan before I approve", "audit what you just did", "ultrathink review". Do NOT trigger for artifact grading alone (use /evaluate), fact-checking drafts (use /validate), or strategic reasoning (use /deep-think). Usage: /scrutinize [plan | execution | file:<path> | dir:<path> | workspace | trajectory:<run_id>] [--relentless] [--no-refute] [--include-low-confidence] [--include-ambiguous]

Second-opinion advisor. Consults Gemini, Grok, and Kimi in parallel for independent views on hard or high-stakes calls. Two modes: independent (all models see the problem only, reason fresh) and critique (all stress-test a draft). Distinct from /deep-think (Claude reasoning structured, alone) and /odin (Claude + the curated knowledge brain). Trigger when the user says: "council", "/council", "second opinion on", "consult the council", "what would Gemini/Grok/Kimi say about", "stress-test this with Gemini/Grok/Kimi", "council vote".

Reflective memory consolidation - performs a periodic pass over memory files, synthesizing recent learnings into durable, well-organized memories for future sessions. Validates technical claims against Context7, enforces security protocols, prunes stale entries, and produces a structured consolidation report. Use when the user says "dream", "/dream", "consolidate memories", "memory cleanup", "reflect on recent sessions", "update memories", or at the end of a productive session when significant new information was learned. Do NOT use for: simple memory writes (use auto-memory), knowledge base operations (use /zk), or session initialization (use /prime).

Odin - personal AI advisor with persistent knowledge brain. Ingests books, articles, videos, documents. Builds principles and positions. Gives referenced advice prioritizing learned knowledge. Nine modes: learn (ingest material), consult (give advice), reflect (review brain), recall (query brain), teach (direct learning), log (record an episode - something that happened), collect (scan business threads + captured comms for episodes you forgot to log), compile (unified linting across brain and ZK), skill-proposal (propose a reflection-derived how-to principle as a checklist step in a target skill - proposal only). Triggers on "Odin" as name/address, "ask Odin", "what would Odin say", "/odin". DO NOT trigger on "odin" as the Russian numeral "one" in phrases like "odin variant" or "another one". Only trigger when used as a proper name or form of address.

Workspace-wide semantic recall. Searches the local associative-memory index (scripts/memory-index.py: bge-m3 hybrid dense+BM25, on-machine, zero API cost) across everything meaningful in the workspace -- Odin brain, business threads, CRM, LinkedIn, context, reference, plans, deliverable outputs, knowledge, datastore extracts, and the auto-memory files -- then answers ONLY from the retrieved sources with file-path citations, or says "not in memory" when the index reports a gap. Use when the user asks "what do we know about X", "where did we decide Y", "have we touched Z before", "recall ...", "search my memory for ...", "find what we said about ...", or wants to surface a past decision / brief / contact / note by meaning rather than exact words. Do NOT use for: Odin-brain-only advice or episode dedup (use /odin recall, which is brain-scoped); external/world intelligence on a company or person (use /osint); capturing a NEW note (use /zk); plain exact-string file search (use Grep). This skill never fabricates be

Compose a short message and deposit it into the Action Queue as a GATED draft card that waits for an explicit human approve before it can ever send. Use to stage an outbound draft (e.g. a demo email) for later review, and as the reference draft-tier skill the headless runner (`heading skill queue-draft`) exercises. NEVER auto-trigger - explicit `/queue-draft` (or an explicit headless run) only. It deposits a draft; it NEVER approves, sends, or calls a send transport.

Execute an approved implementation plan step by step, with optional trajectory logging and a post-run evaluation pass. Use to build out a plan produced by /create-plan. Trigger when the user says "implement", "execute the plan", or "build it". Do NOT use for planning itself (use /create-plan).

Global, geopolitical, and cross-domain intelligence brief for the CEO - world events, sovereign-tech shifts, macro and threat signals relevant to 31C, rendered as a dated HTML/PDF brief. Use for world-scale situational awareness, not a single entity. Trigger when the user says "world intel", "geopolitical brief", "what's happening globally", "global threats", or "CEO intelligence brief". Do NOT use for a specific company or person (use /osint) or an externally published newsletter (use /intel-briefing-newsletter).

Competitive analysis of a named vendor versus 31C/ODUN.ONE - product gaps, geographic overlap, honest strengths, exploitable weaknesses, win strategy, and recommended actions. Use when comparing 31C against a specific competitor or sizing a competitive landscape. Trigger when the user says "competitor analysis", "how does [company] compare to [competitor]", "competitive advantage vs [competitor]", or "competitive landscape for [sector]". Do NOT use when the target is a person (use /osint), for market sizing (use /market-brief), or a generic "vs" with no named second party (ask to disambiguate first).

Personal CRM - add, log, radar, find, update contacts

Produce an external-facing intelligence briefing newsletter - curated, publishable intel framed for an audience beyond the CEO. Use when the output is meant to be shared/published, not a private desk brief. Trigger when the user says "newsletter", "intel briefing", "publish intelligence brief", or "external intel brief". Do NOT use for an internal-only situational brief (use /ceo-intel).

Produce a deep pre-meeting brief on an external counterpart - OSINT dossier, CRM relationship history, prior comms, a Voss tactical-empathy plan, and three calibrated questions. Use before any meeting with a named external person and/or company. Trigger when the user says "meeting prep for [counterpart]", "prepare for meeting with [person or company]", or "briefing for [named person + company]". Do NOT use for internal syncs without an external counterpart (just answer), generic briefings without a named target (use /market-brief or /dashboard); deep multi-signal prep escalates to the orchestrator's Deep Meeting Prep pattern.

EXPLICIT INVOCATION ONLY -- never auto-trigger. This skill activates ONLY when the user literally types '/osint-advanced' or says 'run osint-advanced'. DO NOT trigger from 'investigate', 'research', 'dig into', 'OSINT', 'intelligence on', 'background check', 'due diligence', 'profile', 'recon', 'dossier', 'deep dive', or ANY general intelligence request -- those belong to /osint. This skill queries specialized OSINT databases (sanctions lists, corporate registries, breach databases, threat actor platforms, infrastructure scanners) that the base /osint does not cover. Requires explicit user approval before each execution.

Deep OSINT intelligence gathering on any target - company, person, market/region, or technology. The master intelligence skill that feeds all other intel skills. Produces executive-grade intelligence briefs with confidence ratings, source attribution, and 31C business relevance. Use when the user says 'OSINT', 'investigate', 'dig into', 'deep dive on', 'intelligence on', 'find out everything about', 'who is', 'what do we know about', 'research [company/person/market/tech]', 'background check', 'due diligence on', 'profile [target]', 'recon on', 'intel on', 'dossier on', or asks for comprehensive intelligence gathering on any entity, person, region, or technology. This is the raw intelligence foundation that /competitor-intel, /deal-strategy, /meeting-prep, and /market-brief consume. Always trigger when the user needs deep, comprehensive, multi-source intelligence on a specific target.

Automate web browsers with Playwright for screenshots, scraping, form filling, PDF generation, YouTube video analysis, and website testing. Use for any task requiring a real browser or YouTube content understanding -- extracting data from JavaScript-rendered pages, capturing screenshots of competitor websites, filling and submitting web forms, generating PDFs from web pages, monitoring website availability, testing 31C web properties, or understanding YouTube video content. Trigger when the user says "playwright", "screenshot this site", "scrape this page", "fill this form", "capture this website", "extract data from", "web automation", "browser automation", "PDF from website", "monitor this site", "watch this video", "what's in this YouTube video", "understand this video", "youtube transcript", or any task involving programmatic browser interaction or YouTube content extraction. Also trigger when a URL points to a JS-rendered page that WebFetch cannot handle. Do NOT trigger for simple URL fetching (use WebFe

Generate and edit presentation slides as PPTX files. Also create LinkedIn carousels and manage reusable slide layouts. TRIGGERS - Use this skill when user says: - "create slides for [brand]" / "generate presentation for [brand]" / "make slides for [brand]" - "create a carousel for [brand]" / "linkedin carousel" / "make a carousel about [topic]" - "edit this pptx" / "update the slides" / "modify this presentation" - "create a new layout" / "add a layout to the cookbook" / "make a [type] layout template" - "edit the [name] layout" / "update the cookbook" / "improve the [name] template" - Any request mentioning slides, presentations, carousels, PPTX, or layouts with a brand name Creates .pptx files compatible with PowerPoint, Google Slides, and Keynote. Creates PDF carousels for LinkedIn (square 1:1 format).

Request a new skill from the CEO - describe what it should do, and the request is emailed to the admin. EXPLICIT INVOCATION ONLY - sends external email.

Use this skill to communicate via Telegram as Misha's personal account. Handles all Telegram messaging: send messages to contacts or groups, read conversations, check unread, search message history, forward messages between chats, reply to specific messages, send files/photos, download media, list chats/groups, view contacts, and mark as read. This is a live Telegram client that sends and receives real messages on Misha's behalf. Trigger for any request to interact with Telegram as a user: sending, reading, searching, forwarding, downloading, or managing Telegram conversations and media. Also triggers on "/telegram". Do NOT trigger for building Telegram bots, writing Telegram API code, debugging Telethon scripts, setting up Telegram integrations or MCP servers, or any development task that merely mentions Telegram. Do not trigger for other messaging platforms (Slack, WhatsApp, Signal, email) or generic "send a message" requests without Telegram.

Domain-agnostic YouTube intelligence scanner. Searches YouTube for videos in any topic area (AI, DPI, cybersecurity, markets, sailing, etc.), identifies trending topics, ranks videos by engagement quality, deep-analyzes the top video with full transcript extraction, and delivers an actionable intelligence brief in MD, HTML, and PDF formats. Automates daily YouTube scanning. Use when the user says "yt-pulse", "youtube pulse", "scan YouTube for", "what's trending on YouTube about", "YouTube intel on", "find YouTube videos about", "what are creators talking about", "youtube scan", or asks for YouTube content discovery and analysis in any domain.

Backup the entire workspace to GitHub. EXPLICIT INVOCATION ONLY - never auto-trigger from natural language.

Reflective end-of-session self-improvement. Scans the current Claude Code session for corrections, preferences, repeated patterns, errors, success patterns, and voice violations, then proposes numbered concrete patches to memory, settings, ceo-only skills, and ceo-only rules. Corporate files route to a separate review queue and are NEVER auto-applied. Use at end of every working session. Light mode for low-token state or quick sweeps. CEO-only - never propagates to execs.

Pre-implementation gate: recommended structured checkpoint between plan approval and /implement. Use before any non-trivial implementation: combines success-criteria definition, an inline contrarian stress-test (the /devil discipline), an optional /council architecture review with Kimi as devil's advocate, harness gap audit, and test-contract writing. Produces a GO/NO-GO artifact with a ready-to-paste /implement prompt. Skip for trivial one-liner fixes, typo corrections, and config-only changes.

CEO-only: publish workspace content to all executives via the corporate repo. EXPLICIT INVOCATION ONLY - never auto-trigger.

CEO-only: Push all workspace updates to all executives. Single command that commits, classifies, publishes to corporate repo, bumps BUILD.json, pushes CRM, aggregates CRM, syncs exec workspaces, and reports. EXPLICIT INVOCATION ONLY - never auto-trigger.

Manage the Sentinel background daemon that monitors corporate email (Exchange) and Telegram for urgent messages. Start, stop, check status, view logs, and configure monitoring. EXPLICIT INVOCATION ONLY - daemons live on the always-on service host and are off-limits without CEO sign-off.

Interactive setup wizard for HEADING OS and 31C exec workspaces. Walks the user through ~22 questions (public) or ~9 questions (exec), enriches short answers into rich docs, captures API keys securely, and produces a completion-% dashboard on re-runs. Refuses to run on the CEO master workspace. EXPLICIT INVOCATION ONLY - mutates workspace skeleton (.env, reference/, context/, personal/).

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. EXPLICIT INVOCATION ONLY - mutates workspace skill infrastructure.

Manually trigger workspace sync - git pull of the code/data clones, then a push-all backup. EXPLICIT INVOCATION ONLY - never auto-trigger.

Draft peer-level CEO-to-CEO correspondence in Misha's voice - executive register, relationship-aware, Voss-calibrated. Use for writing to another company's CEO or founder as a peer. Trigger when the user says "CEO letter", "write to [CEO name]", "peer correspondence", or "executive letter". Do NOT use for a non-CEO recipient (use /email-draft) or a formal branded external letter on letterhead (use /corporate-letter). NEVER auto-sends - produces a draft for review.

Drain overdue CRM "red debt" into the Action Queue as routed, voice-drafted nudges for one-click CEO approval. Use when the CEO says "cold sweep", "drain the cold contacts", "sweep the overdue CRM", or wants the overdue-contact backlog turned into ready-to-send drafts. NOT for sending email (that is the human-approved executor) - this skill only drafts into the queue. NOT for a single follow-up (use /follow-up). NOT for pipeline review (use /crm). CEO-only (not synced to executives) during the prove-out.

Fetch up-to-date, version-specific library documentation from Context7 for accurate API references, code examples, and usage patterns. Use when working with any library or framework and you need current documentation instead of relying on training data. Trigger when the user says "context7", "/context7", "look up docs for [library]", "get documentation for [library]", or when you need to validate code against current library APIs. Also use proactively when writing code that depends on external libraries to ensure correctness.

Design a structured implementation plan for a non-trivial change before any code is written - objective, scope, critical files, step sequence, success criteria, and risks. Use before significant additions or multi-file work. Trigger when the user says "create plan", "plan for [change]", or "design the approach". Do NOT use to execute an existing plan (use /implement).

Generate the daily CEO Morning Dashboard -- a single-page operational briefing aggregating CRM health, pipeline, calendar, email, strategy, and data freshness.

Prepare investor data room documents, due diligence responses, and fundraising materials. Use when preparing for investor conversations, responding to due diligence questions, or building the formal data room. Trigger when the user says "data room", "due diligence", "DD response", "investor materials", or asks to prepare documents for investors.

Full deal strategy analysis for a specific prospect or opportunity. Produces prospect intelligence, competitive positioning, pricing recommendation, objection handling, and next steps. Use when evaluating a new deal, preparing for a sales engagement, or building competitive positioning. Trigger when the user says "deal strategy", "analyze this deal", "how do we win this", "competitive positioning for [prospect]", or asks to evaluate a prospect or opportunity.

Professional design system - illustrations, UI mockups, social graphics, brochures, logos, infographics, and image editing. Routes to optimal pipeline (HTML Studio for layouts, Replicate AI for creative imagery, PPTX for presentations). Use when user asks to design, create visuals, generate graphics, make a mockup, create a brochure, design a logo, or produce any visual asset. Do NOT use for simple one-off image generation (use /flux-image instead).

Parse documents (PDF, DOCX, PPTX, XLSX) with spatial bounding boxes via LiteParse. Answer questions with visual citations showing exact source locations on page images. Generate 31C-branded HTML/PDF reports. Maintains a parse cache for efficiency. Use when: "parse this document", "docparse", "visual citations", "show me where it says", "document analysis with citations", "parse with bounding boxes", "extract from this PDF". NOT for: plain text extraction without spatial data (use datastore-extract.py), email analysis (use /email-intel), web scraping (use /playwright or /firecrawl).

Draft a reply to an existing email thread in Misha's voice - reads the incoming message, mirrors context, and answers with a Voss-calibrated response. Use when responding to a received email. Trigger when the user says "respond to this email", "reply to this", or "draft reply". Do NOT use for a new outbound email (use /email-draft). NEVER auto-sends - produces a draft for review.

Skeptical quality evaluator for workspace artifacts. Grades skills, scripts, reference files, and rules against workspace standards and plan success criteria. Runs deterministic checks via scripts/artifact-evaluator.py, then applies qualitative judgment on clarity, craft, and completeness. Use after /implement or standalone on any file path. Produces evaluation report with specific, actionable feedback. Triggers on: "evaluate", "grade", "review quality", "check this artifact", "evaluate this skill/script".