nemock

Close the loop after a hot-sheet recording session. Takes the episode number and which cards the host actually riffed on (the hot sheet page's "Copy riff summary" button produces the exact command), marks the source inbox nodes as acted on so upstream pipelines stop resurfacing them, and — when a transcript or recording notes are provided — extracts quotable lines, fresh positions, and anecdotes into a handoff file for the voice-library intake. Run this skill when the user says "riff-capture EP### --cards 1,3,5", "capture that session", "I recorded, here's what I covered", "close out the episode", or pastes the copied riff summary. This is the downstream counterpart to hot-sheet.

Build a glanceable, browser-based hot sheet for a solo riff-to-camera episode and open it in a browser tab. Reads topic candidates from configured local inboxes (research-ingest nodes, voice-library positions) and/or the latest topic-radar shortlist, picks a themed set of riffable items, drafts a cold-open intro and per-topic cards with context, source links, the host's pre-loaded angle, and riff prompts, then renders a single self-contained HTML page via scripts/render_hotsheet.py. Run this skill when the user says "let's do an episode", "build a hot sheet", "hot sheet", "set up a riff session", "I want to record", "what am I riffing on today", or any equivalent request to get camera-ready fast. This is the solo-format counterpart to episode-program; riff-capture closes the loop afterward.

Scan an episode transcript for short-form clip candidates suitable for Instagram Reels, TikTok, and YouTube Shorts. Targets punchy openers, contrarian takes, surprising stats, emotional beats, and complete narrative arcs that stand alone without surrounding context. Produces 8–12 candidates per episode with start and end timestamps, a transcript excerpt, a suggested hook caption, a platform recommendation, and a why-this-clips note. Outputs a markdown file for the host to scan and a JSON file an editor can import. Run this skill when the user says "find clips for EP042", "what's clippable", "pull clips from the recording", "find shareable moments", "build the clip list", "clip finder", or any equivalent ask for short-form moments. Run after show-notes (the cleaned transcript is the preferred input).

Build a separate prep document for the remote co-host of an upcoming episode. Reads the run-of-show and any research briefs in the episode folder, then drafts a co-host-facing artifact in the right framing — cold-open setup line, segment-by-segment "throw" lines that hand off to the host, two to four follow-up questions per segment, articles and links the co-host should have queued, pacing notes (when to interject, when to let the host run uninterrupted), and recurring segment cues. Outputs both a local markdown file and a shareable Google Doc when Drive is connected. Run this skill when the user says "build the co-host brief", "prep the co-host", "make the co-host doc", "send the co-host their prep", "cohost brief for EP042", or any equivalent ask for the second-seat prep document. Run after episode-program, ideally after research-brief has been run for each topic.

Surface themes that span multiple past episodes and suggest concrete callback opportunities for the next episode. Reads cleaned transcripts and topic shortlists across the recent run, clusters recurring themes (e.g., "AI commoditization", "Spotify ad strategy", "founder talent flight"), and produces a callbacks document that names the theme, when it last came up, what was said, and a suggested phrasing the host can drop in. Also writes a longer themes map that catalogs every running thread across the lookback window. Run this skill when the user says "what arcs are running across episodes", "find callbacks for EP043", "cross-episode themes", "what should I call back to", "callback opportunities", "what are the running threads", or any equivalent ask about story continuity. Run after episode-program for the next episode is drafted but before recording, so the host can absorb the callbacks during prep.

Build the social distribution package for a published episode. Produces a LinkedIn post in the host's voice (1300–2000 characters), an X/Twitter single post plus a 4-tweet thread option, an Instagram caption built around the featured clip, a YouTube community post, and an optional newsletter tie-in section when the host runs a tie-in newsletter. Voice-checked against banned words and banned phrases. Schedules posts via Blotato when the connector is available; falls back to writing drafts to a local file. Run this skill when the user says "build the distribution pack", "draft the social posts", "social copy for EP042", "promote EP042", "make the LinkedIn post", "schedule the posts", "distribution pack", or any equivalent ask for episode promotion. Run after show-notes and clip-finder.

Maintain the show's editorial calendar in the configured Google Sheet (or local CSV fallback) and sync recording and publish dates to the configured Google Calendar so the co-host sees the schedule. Supports adding new episodes, updating status and metadata (recorded, edited, published, downloads, top clip), querying single episodes, and a "what's on deck" report that lists upcoming episodes and what's missing for each. Run this skill when the user says "update the editorial calendar", "what's on deck", "next episode status", "mark EP042 as published", "schedule EP043 recording", "what's missing for EP043", "sync the calendar", "add EP043 to the calendar", or any equivalent ask about episode status, scheduling, or the calendar pipeline. The default action with no arguments is the next-episode report.

Build a run-of-show for the next episode from two or three selected topics. Reads the most recent topic-radar shortlist (or accepts free-form topic descriptions if topic-radar has not been run), allocates time across recurring segments and topic segments using the show's configured runtime, and drafts a structured rundown with cold-open hook candidates, host talking-point bullets, co-host setup lines, follow-up questions, transitions, and a closing call-to-action. Writes a markdown file locally and creates a shareable Google Doc when the Drive connector is available. Run this skill when the user says "build a run-of-show", "create the rundown", "plan the next episode", "draft an episode outline", "make a run of show", "rundown for episode X", "episode-program", or any equivalent request to structure the next recording. This is the upstream input for research-brief and cohost-brief.

Interactive Q&A capture of company knowledge into typed nodes in a company-brain vault. Sub-modes - vision (dictation-friendly six-phase flow), for-doc (document-driven interview that walks a target doc's sections), product, persona, competitor, competitor-ifu, competitor-clearance, competitor-snapshot, metric, meeting-notes, risk, clearance. Each session writes new typed markdown node files into the right folder and links them via derived_from edges to a source node. Always reads the vault's _system/PROFILE.md first to learn the active schema, never assumes it.

Generate planning documents from a company-brain vault. Ships 21 generators: full MRD (markdown / html / docx) and one-pager (markdown / html); plus 19 scaffolds for PID, project-charter, stakeholder-register, risk-register, status-report, meeting-minutes, lessons-learned, business-plan, sales-battle-card, competitive-brief, ifu-comparison, decision-log, press-release, investor-update, onboarding-doc, srd, srs, hrs, risk-brainstorm. Profile-aware — medical-device-only docs (risk-register, ifu-comparison, risk-brainstorm) reject non-medical-device vaults. All output is idempotent and lands under <vault>/exports/.

Audit and repair a company-brain vault. cb maintain repair auto-fixes filename-id mismatch, missing inverse edges (preceded_by ↔ followed_by), and missing controlled_document:false on risk/IFU nodes; it also regenerates _system/INDEX.md and refreshes the auto-section of the vault-level README.md (between cb:auto markers). cb maintain decay applies half-life confidence decay to fact snapshots linked to volatile metrics. cb maintain rebuild-index regenerates _system/INDEX.md only. cb maintain rebuild-readme regenerates the README auto-section only. cb maintain init-readme-markers and cb maintain init-gitignore-markers add the cb:auto and cb:gitignore-managed markers to legacy vaults (non-destructive upgrade path). cb maintain audit reports vault health read-only. cb validate --fix wires the repair pass into validation.

Generate a D3-based interactive HTML viewer for a company-brain vault. Single self-contained file the user opens in any browser — no server. Three view modes: graph (default, full force-directed graph, color by node type), ifu-chain (indication-for-use nodes connected via preceded_by/followed_by, medical-device only), predicate-tree (regulatory-clearance nodes connected via preceded_by, medical-device only). Branding palette and template overrides supported.

Answer questions against a company-brain vault using staged retrieval, auto-injected pillars, typed edge traversal, and node-id citations. Profile-aware. Cites the graph as the source of truth; flags contradictions, staleness, and confidence.

Ingest existing documents (markdown, plain text, Word .docx, PDF, interview transcripts, image screenshots) into typed company-brain nodes with provenance. Each ingested document lands as an immutable source node; extracted content becomes derived typed nodes linked via derived_from edges. Uses `cb extract` for binary documents; uses Claude's native vision for image screenshots; reads markdown and plain text directly. Recognizes known skill-output structures (competitor-profiling, customer-research, write-spec, etc.) and routes content accordingly.

Scaffold a new company-brain vault. Use ONCE when starting a new vault. Confirms the target directory, industry profile, and whether to initialize git (default yes), then runs `cb scaffold` to create the folder tree, _branding/, _attachments/, exports/, and the _system reference files (PROFILE.md, INDEX.md, NODE-TYPES.md, EDGE-TYPES.md, FRONTMATTER-SCHEMA.md). Initializes the vault as a git repository by default. Does NOT write node content — that is the job of intake and atomize.

Compile every press release ever issued by a public company into a structured local markdown archive (per-release files + a top-level INDEX.md). Use when the user wants to "archive press releases for [Company]", "study [Company]'s PR strategy", "build a corpus of [Company]'s announcements", or wants offline-resilient access to a competitor's wire history. Works best on US-listed companies (SEC EDGAR is the most reliable source); pre-IPO archives are sparser. Requires curl + python3. Optional Chrome Connector (Claude Desktop) for releases that aren't in EDGAR or Wayback Machine.

Reverse-engineer the PR strategy of a public company from a structured press-release archive (the output of press-release-archiver). Produces a strategic playbook with credibility-ladder analysis, drumbeat patterns, stakeholder orchestration, and foundation-phase plays — designed for early-stage companies (seed → Series A → pre-IPO) who want to learn from how mature companies built their narrative. Supports single-company analysis OR comparative analysis with stage alignment (e.g., "what was Globex Corp doing in its first 5 years post-IPO, vs. Acme Medtech in its first 5?"). Token-conscious: deterministic Python computes stats, then YOU (Claude, in this conversation) synthesize the strategic narrative directly.

Turn a strategic PR-analysis playbook (the output of press-release-analyzer) into a recording-optimized HTML slide presentation, ending with concrete recommendations for the client's own PR strategy. Single self-contained HTML file with keyboard navigation, dark theme, big numbers, comparison layouts, timelines, and chapter sidebar. The closing chapter is always client-facing recommendations the user can hand to a seed → Series A → pre-IPO company. Sibling skill to press-release-archiver and press-release-analyzer.

Produce a long-form (~20+ min, 16:9) operator-narrated deep-dive video — an orchestrated film of many small `explainer` segments (cold-open → Act I → sponsor → Act II → sponsor → Act III → CTA), assembled RAM-safely into one master with chapters + captions. Use when the user wants a "deep dive", "long-form explainer", "YouTube tutorial video", "20-minute video", or runs "/deepdive <topic>". Brand-parameterized (FWF-first). You author the plan + per-segment script/deck and drive the record/review loop; a pure-Python pipeline does narrate→align→render →mux per segment and conform→concat→validate for the master. Generation only — it writes a labeled program dir + crash-safe manifest; it NEVER posts to social platforms (that's Phase 3).

Turn a topic (or source document) into a visually dynamic HTML explainer deck and a narrated vertical video, end-to-end, using only local/free tools (Kokoro TTS, torchaudio forced alignment, Playwright, ffmpeg) plus this Claude session. Use when the user wants to "make an explainer video", "turn this into a Short/ Reel/TikTok", "create an explainer deck", or "/explainer <topic>". Supports topic-only OR source-driven (ingest a PDF/URL and frame a real figure/screenshot); aspects 9:16, 16:9, 4:5; fixed theme. Generation only — it writes a labeled output dir + manifest.json; it does NOT post to social platforms.

Produce a YouTube-intelligence-driven, retention-engineered explainer video (deep dive, Shorts, a masterclass/Operator's-Guide series episode, or a promotional video) end-to-end on this Mac: intel sweep → Blueprint → script → operator recording (or Kokoro) → align → render → package. Use when the user says "/explainer-studio <topic>", "make a video about", "run the studio pipeline", or asks for a Blueprint on a topic. Local/free only; the single allowed subscription is the operator's stock.adobe.com membership, human-in-the-loop. This skill NEVER posts to social platforms.

Pressure-test any project, system, document, or plan from multiple independent expert perspectives at once, then synthesize one prioritized report. Convenes a "council" of reviewer agents — each assigned a distinct lens (e.g. brand/voice, content/editorial, technical reliability, distribution/growth, operations/ maintainability) — that review in parallel, then a chair synthesizes their findings, surfacing agreements and conflicts. Use when the user asks for a "council review," "multi-perspective review," "pressure test," "review this from different angles," "panel review," "get multiple perspectives on," "red-team this," or wants a project analyzed from several points of view. Report-only by default — it recommends, it does not change anything.

Use this skill whenever the user wants to clean up, fix, master, denoise, de-hum, level, normalize, or enhance the audio on a video file or audio file. Trigger on direct requests like "clean up the audio," "fix this audio," "master the audio," "make this sound like a podcast," "podcast-grade audio," "enhance my voice," "remove the hum," "remove background noise," "denoise," "de-hiss," "the audio sounds bad," "voice sounds muddy/tinny/quiet/uneven," "even out the levels," "make it broadcast loud," "normalize loudness to -16 LUFS," "remove 60 Hz hum" (or 50 Hz), or any indirect ask where they share a recording and want it to sound professional. Works on MP4, MOV, MKV, WAV, MP3, M4A, AAC. Preserves the video stream untouched.