@lifeaitools/rdc-skills 0.25.4 → 0.25.9

package/skills/channel-formatter/SKILL.md

@@ -1,538 +1,538 @@
----
-name: rdc:channel-formatter
-description: "Usage `rdc:channel-formatter <channel|pack> [content]` — Apply precise, channel-native formatting and content repurposing to any output: LinkedIn, Twitter/X, Slack/Teams, Email (external/internal), Pitch Deck slides, Word/DOCX, PDF Report, Web/Landing Page, and multi-output content packs. Use EVERY TIME the user names an output channel/platform/document type, asks to 'write a post', 'draft an email', 'format this for', 'make this a slide', 'send to LinkedIn', 'write a tweet', 'social media post', 'turn this article into posts', 'make a content pack', 'make a social pack', or 'repurpose this article'. Each channel has its own Unicode strategy, emphasis system, length limits, and structure — never apply generic markdown to channel-specific output. This skill FORMATS, STRUCTURES, and REPURPOSES text only — for actual .docx/.pptx ↔ Markdown FILE conversion (either direction) use `rdc:convert` (build-corpus), not this skill. Self-contained: all channel and pack rules are inlined below (no external reference files)."
----
-
-> **⚠️ OUTPUT CONTRACT (READ FIRST):** `guides/output-contract.md`
-> The formatted deliverable IS the output. Produce the channel-correct text directly.
-> No tool-call narration, no raw markdown wrappers around channel output, no log dumps.
-
-> If dispatching subagents or running as a subagent: read `{PROJECT_ROOT}/.rdc/guides/agent-bootstrap.md` first.
-
-> **Sandbox contract:** This skill honors `RDC_TEST=1` per `guides/agent-bootstrap.md` § RDC_TEST Sandbox Contract. It writes no files and makes no external calls; under `RDC_TEST=1` behavior is unchanged (pure text transform).
-
-# Channel Formatter
-
-Format or repurpose output precisely for the target channel. This skill is
-**self-contained** — every channel and pack rule is inlined below. Detect the
-channel or pack, jump to its section, apply its rules exactly. Never apply
-generic markdown to a channel that doesn't render it.
-
----
-
-## When to Use
-
-Use this skill whenever the user asks to write, format, adapt, or repurpose
-content for a specific channel, platform, document type, or output pack.
-
-Common triggers:
-- "write a LinkedIn post"
-- "turn this article into social posts"
-- "make a content pack"
-- "make a social pack"
-- "repurpose this article"
-- "write a tweet/thread"
-- "draft an email"
-- "format this for Slack"
-- "make this a slide"
-- "write landing-page copy"
-
-Do not use this skill for binary file conversion, PDF rendering, Brochurify
-orchestration, or brochure JSX authoring; delegate those to the specialist
-skills named in the scope boundary.
-
-## Arguments
-
-`rdc:channel-formatter <channel|pack> [content]`
-
-- `<channel|pack>`: optional target such as `linkedin`, `twitter`,
-  `twitter-thread`, `slack`, `email-ext`, `email-int`, `pitch-deck`, `word`,
-  `pdf-report`, `web`, `strict-format`, `social-pack`, `campaign-pack`,
-  `exec-pack`, or `launch-pack`.
-- `[content]`: optional source text, path reference, article, report, transcript,
-  rough draft, or surrounding conversation content.
-
-## Channel Detection
-
-| If user says / context is...                          | Channel          | Go to section            |
-|-------------------------------------------------------|------------------|--------------------------|
-| LinkedIn, post, share on LinkedIn                     | LinkedIn         | [§ LinkedIn](#-linkedin) |
-| Tweet, Twitter, X post, thread                        | Twitter/X        | [§ Twitter / X](#-twitter--x) |
-| Slack, Teams, Discord, chat message                   | Slack/Teams      | [§ Slack / Teams](#-slack--teams) |
-| Email to client/investor/external, cold outreach      | Email (External) | [§ Email — External](#-email--external) |
-| Email to team, internal, Slack alternative            | Email (Internal) | [§ Email — Internal](#-email--internal) |
-| Pitch deck, slide, investor slide, one-pager slide    | Pitch Deck       | [§ Pitch Deck](#-pitch-deck) |
-| Word, DOCX, document, report, memo, letter, contract  | Word/DOCX        | [§ Word / DOCX](#-word--docx) |
-| PDF report, annual report, white paper                | PDF Report       | [§ PDF Report](#-pdf-report) |
-| Website, landing page, web copy                       | Web/Landing Page | [§ Web / Landing Page](#-web--landing-page) |
-| Artifact, JSX, React component, Claude artifact       | Artifact/JSX     | → use the `jsx-author` / `impeccable` skill instead |
-
-### Pack Detection
-
-| If user says / context is...                          | Mode             | Go to section                 |
-|-------------------------------------------------------|------------------|-------------------------------|
-| social pack, content pack, posts from this, repurpose this article for social | `social-pack` | [§ Content Repurposing Packs](#-content-repurposing-packs) |
-| campaign pack, launch campaign, article into email + social + web | `campaign-pack` | [§ Content Repurposing Packs](#-content-repurposing-packs) |
-| exec pack, executive pack, leadership summary, internal briefing | `exec-pack` | [§ Content Repurposing Packs](#-content-repurposing-packs) |
-| launch pack, announce this, launch posts              | `launch-pack`    | [§ Content Repurposing Packs](#-content-repurposing-packs) |
-| strict format, preserve wording, do not rewrite       | `strict-format`  | [§ Repurposing Modes](#-repurposing-modes) |
-
----
-
-## Workflow
-
-1. **Detect channel or pack mode** from the request using the tables above.
-2. **Classify the source**: already-drafted copy, long article/report, transcript, notes, or brief.
-3. **For thin or under-specified long sources**, enrich before writing instead
-   of inventing context:
-   - If the source names a corpus path, read it.
-   - Otherwise search/read the approved corpus first when available. Resolve
-     `CORPUS_ROOT` (usually `H:/My Drive/global-corpus`) and
-     `LOCAL_CORPUS_ROOT` (usually `C:/Dev/local-corpus`) before searching the
-     repo. If environment variables are not visible, try those default paths.
-   - The corpus search must use an explicit corpus path in the tool arguments
-     (`H:/My Drive/global-corpus` or `C:/Dev/local-corpus`). A relative repo
-     search does not count as enrichment.
-   - Use `Grep`/`Glob`/`Read` or the environment's equivalent tools so the
-     transcript records what was consulted.
-   - Use web search only when the requested output needs current/public facts,
-     the user asks for external context, or corpus context is absent and network
-     search is allowed. Cite/use only what the search actually supports.
-   - Searching only the project repo is not enough for enrichment unless the
-     source itself points to a repo-local corpus file.
-   - If neither corpus nor web context is available, keep the output sparse and
-     mark missing context in assumptions; do not fill gaps creatively.
-4. **For long sources**, extract thesis, audience, proof points, CTA, constraints, and factual risks before writing.
-5. **Jump to the target channel or pack section** below and apply all rules exactly — do not rely on memory.
-6. **Never mix** markdown conventions across channels.
-7. If the channel or pack is ambiguous, ask once: "Is this for [Channel A] or [Channel B]?"
-8. Produce the formatted or repurposed output directly as the deliverable.
-9. Treat extraction notes and source-fidelity guardrails as internal scratchwork.
-   Do not print a "Long-Source Extraction Checklist", factual guardrail list, or
-   absent-topic list unless the user explicitly asks for analysis.
-
-## Hard Rules (all channels)
-
-- Never apply generic markdown (`##`, `**`, `-`, etc.) to LinkedIn, Email, Slack, or Twitter output.
-- Never use raw LaTeX in LinkedIn, Email, Slack, or Twitter — use Unicode math symbols.
-- Never use Word Styles notation in plain-text channels.
-- Always match the tone register of the channel (formal ≠ casual ≠ punchy).
-- For LIFEAI/PRT/RDC content, maintain REGEN-MODE voice unless Author-Mode is active.
-- When repurposing a long source, preserve the source thesis and never invent unsupported statistics, quotes, names, dates, citations, results, or commitments.
-- If the source lacks a needed CTA, audience, proof point, or date, either use a clearly generic placeholder or state the assumption before the polished output.
-
-> ## ⛔ Scope boundary — this skill FORMATS text; it does NOT convert files
-> This skill governs **how to structure and format content** for a channel. It never
-> reads or writes binary office files. For the actual **Word ↔ Markdown (and .pptx/.ppt)
-> file conversion — in either direction — use `build-corpus` / the `rdc:convert` skill**,
-> never this one. So: "convert this .docx to markdown" / "turn this markdown into a Word
-> doc" → **`rdc:convert`**. "format this content the Word way / write it for LinkedIn" →
-> this skill. The Word/DOCX and PDF sections below describe target structure only; producing
-> the actual `.docx`/`.pdf` artifact is `rdc:convert` / `rdc:brochure`, not channel-formatter.
->
-> **Work in Markdown, materialize at the end.** Draft and format everything as Markdown
-> ("MD time" — cheap tokens, clean diffs); only hand off to `rdc:convert` to emit the
-> `.docx`/`.pptx` as the last step. build-corpus round-trips losslessly, so this costs
-> no fidelity. See the build-corpus agent primer (`AGENTS.md`).
->
-> Specialist routing:
-> - Office/Markdown file conversion → `rdc:convert`
-> - HTML/folder/zip/URL to PDF brochure rendering → `rdc:brochure`
-> - Brochurify orchestration jobs → `rdc:brochurify`
-> - Brochure JSX using `@lifeai/brochure-kit` → `lifeai-brochure-author`
-
----
-
-## § Repurposing Modes
-
-Use these modes when the input is longer than the requested output, such as an
-article, memo, transcript, report, rough draft, or notes.
-
-### `strict-format`
-Preserve the source wording and argument as much as the channel allows. Use this
-when the user says "format only", "do not rewrite", "preserve wording", or when
-legal/compliance precision matters. You may adjust line breaks, headings,
-emphasis, bullet symbols, and channel-specific structure, but do not add new
-claims or reframe the argument.
-
-### `single-channel`
-Repurpose the source into one target channel. Extract the usable argument,
-choose the strongest hook for that channel, compress or expand to the channel's
-native length, and produce one finished output.
-
-### `social-pack`
-Repurpose one source into a coordinated set of social outputs:
-- LinkedIn thought-leadership post
-- Short LinkedIn announcement or teaser variant
-- Twitter/X single post
-- Twitter/X thread
-- Slack/Teams internal share
-
-### `campaign-pack`
-Repurpose one source into a small campaign kit:
-- Everything in `social-pack`
-- External email intro/blurb
-- Web excerpt
-- SEO meta title and meta description
-- 3 CTA variants
-
-### `exec-pack`
-Repurpose one source for internal leadership alignment:
-- Internal email summary
-- Slack/Teams update
-- Executive-summary paragraph
-- 3 talking points
-- Decision/ask line if the source supports one
-
-### `launch-pack`
-Repurpose one source for an announcement:
-- LinkedIn launch post
-- Twitter/X launch post
-- Slack/Teams launch note
-- External email blurb
-- Web hero headline/subheadline/CTA
-- 3 CTA variants
-
-### Long-Source Extraction Checklist
-Before writing from a long source, identify internally:
-- **Thesis:** the central argument or announcement
-- **Audience:** who this is for
-- **Proof:** facts, examples, data, names, dates, or quotes explicitly present
-- **CTA:** what the reader should do next
-- **Tone:** formal, conversational, punchy, executive, regenerative, etc.
-- **Constraints:** length, compliance, brand voice, channel quirks
-- **Gaps:** missing proof, missing CTA, unsupported claims, unclear audience
-
-### Source-Fidelity Rules
-- Do not invent statistics, dates, quotes, citations, partnerships, revenue,
-  legal claims, customer names, or outcomes.
-- Do not infer adjacent finance/reporting concepts that sound plausible but are
-  absent from the source, such as pro forma assumptions, reporting cadence,
-  offset mechanisms, governance structures, verification status, or portfolio
-  implications.
-- If corpus or web context was consulted, use it only for facts it directly
-  supports. Do not blend generic domain knowledge into the source as if it were
-  documented.
-- If the source provides a theme label without an explanation, keep the theme
-  label or paraphrase it conservatively. Do not add explanatory clauses that
-  define how it works, who governs it, how it is verified, how often it reports,
-  or what financial model it opposes unless those details are explicit.
-- Do not contrast patient capital with quarters, earnings calls, fund cycles,
-  exits, venture speed, or portfolio management unless those words or concepts
-  appear in the source or consulted corpus/web material.
-- Do not upgrade tentative language into certainty.
-- Do not turn illustrative examples into facts.
-- Preserve caveats when they affect meaning.
-- If a stronger hook needs a proof point the source does not provide, write a
-  proof-neutral hook instead.
-- If the source says a topic is absent, excluded, or not mentioned, treat that
-  as an internal guardrail. Do not repeat the absent topic in public-facing
-  channel copy, assumptions, caveats, notes, headings, or summaries unless the
-  user explicitly asks for a contrast, compliance note, or risk disclosure.
-  Strip absent-topic lists from the deliverable entirely.
-- When assumptions are material, include a short "Assumptions:" line before the
-  deliverable rather than burying uncertainty in polished copy.
-- Do not output your extraction checklist. The deliverable should start with the
-  requested channel/pack output, except for a brief "Assumptions:" line when a
-  material gap affects the copy.
-- If you include an "Assumptions:" line, limit it to missing useful inputs such
-  as org name, audience, location, date, CTA, or link. Never list absent,
-  excluded, or not-mentioned topics in the assumptions line. If the only
-  uncertainty is an absent-topic guardrail, omit the assumptions line.
-
----
-
-## § Content Repurposing Packs
-
-When a pack is requested, label each output clearly and make every item
-channel-native. A pack is not a generic summary repeated in several lengths.
-
-### `social-pack` Output Shape
-1. `LinkedIn thought-leadership post` — 900-1300 characters, hook-first, white space, Unicode emphasis only when useful, up to 3 hashtags.
-2. `LinkedIn short variant` — 400-700 characters, announcement or teaser style.
-3. `Twitter/X single post` — 265 characters or fewer unless the user asks otherwise.
-4. `Twitter/X thread` — 5-7 tweets, each self-contained and below 280 characters.
-5. `Slack/Teams internal share` — 3-6 lines, direct, with a clear FYI/action/decision framing.
-
-### `campaign-pack` Output Shape
-1. Full `social-pack`
-2. `External email intro` — subject line plus 80-150 word blurb with one ask
-3. `Web excerpt` — 80-120 words, scannable, CTA-ready
-4. `Meta title` — 50-60 characters
-5. `Meta description` — 150-160 characters
-6. `CTA variants` — 3 verb-led CTAs
-
-### `exec-pack` Output Shape
-1. `Internal email summary` — subject plus 6-12 lines
-2. `Slack/Teams update` — 3-5 lines
-3. `Executive summary paragraph` — 100-150 words
-4. `Talking points` — 3 concise bullets/fragments
-5. `Decision or ask` — one line, only if supported by source
-
-### `launch-pack` Output Shape
-1. `LinkedIn launch post` — announcement structure
-2. `Twitter/X launch post` — single post
-3. `Slack/Teams launch note` — internal update
-4. `External email blurb` — subject plus short body
-5. `Web hero` — headline, subheadline, CTA
-6. `CTA variants` — 3 options
-
-### Pack Quality Rules
-- Vary hooks by channel; do not repeat the same first sentence everywhere.
-- Keep the source thesis consistent across all outputs.
-- Adapt CTA strength to the channel: softer on thought leadership, direct in
-  email/web, concise in Slack.
-- Use channel-specific formatting rules from the sections below.
-- If source proof is weak, use curiosity and framing instead of inflated claims.
-- Do not include extraction notes, guardrail notes, or absent-topic caveats in
-  the pack. Those are reasoning aids, not channel outputs.
-
----
-
-## § LinkedIn
-
-### Core Principles
-- No markdown whatsoever — LinkedIn renders it as raw symbols.
-- Unicode is the ONLY way to achieve bold, italic, and special emphasis.
-- Emoticons are structural tools, not decoration — use them as bullet replacements and section markers.
-- Hook in line 1 — LinkedIn truncates after ~2 lines before "see more".
-- White space is content — single-line breaks create rhythm and readability.
-
-### Unicode Character Sets
-**Bold (Mathematical Bold)** — key terms, company names, headlines, CTAs:
-```
-𝗔𝗕𝗖𝗗𝗘𝗙𝗚𝗛𝗜𝗝𝗞𝗟𝗠𝗡𝗢𝗣𝗤𝗥𝗦𝗧𝗨𝗩𝗪𝗫𝗬𝗭
-𝗮𝗯𝗰𝗱𝗲𝗳𝗴𝗵𝗶𝗷𝗸𝗹𝗺𝗻𝗼𝗽𝗾𝗿𝘀𝘁𝘂𝘃𝘄𝘅𝘆𝘇
-𝟬𝟭𝟮𝟯𝟰𝟱𝟲𝟳𝟴𝟵
-```
-**Italic (Mathematical Italic)** — emphasis, quotes, sub-themes:
-```
-𝘈𝘉𝘊𝘋𝘌𝘍𝘎𝘏𝘐𝘑𝘒𝘓𝘔𝘕𝘖𝘗𝘘𝘙𝘚𝘛𝘜𝘝𝘞𝘟𝘠𝘡
-𝘢𝘣𝘤𝘥𝘦𝘧𝘨𝘩𝘪𝘫𝘬𝘭𝘮𝘯𝘰𝘱𝘲𝘳𝘴𝘵𝘶𝘷𝘸𝘹𝘺𝘻
-```
-**Bold Italic:**
-```
-𝘼𝘽𝘾𝘿𝙀𝙁𝙂𝙃𝙄𝙅𝙆𝙇𝙈𝙉𝙊𝙋𝙌𝙍𝙎𝙏𝙐𝙑𝙒𝙓𝙔𝙕
-𝙖𝙗𝙘𝙙𝙚𝙛𝙜𝙝𝙞𝙟𝙠𝙡𝙢𝙣𝙤𝙥𝙦𝙧𝙨𝙩𝙪𝙫𝙬𝙭𝙮𝙯
-```
-**Math / Equations** — Unicode symbols, never LaTeX: `× ÷ ± ≠ ≤ ≥ ≈ ∴ ∵ Σ Δ α β ∞ √ ² CO₂ H₂O → ← ⟺ % °`
-Example: instead of `E = mc^2` write `E = mc²`; instead of `\alpha_{PRT}` write `αPRT`.
-
-### Emoticons as Structure
-Section openers / bullets: 🔹 primary · 🔸 secondary · ▸ tertiary · ✅ result · ❌ contrast · ⚡ urgency · 🌱 regenerative (LIFEAI/PRT) · 💡 insight · 📌 takeaway · 🔑 key · 📊 data · 🏗️ development · 🤝 partnership · 💰 capital · 🌍 planetary.
-Dividers: `——————————————` · `▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬` · `· · · · · · · · · · · · · ·`
-
-### Post Structure
-**Standard (thought leadership):** hook line (no period, open loop) → blank → 3-6 short paragraphs (1-3 lines), emoticons as bullets → blank → insight/contrast line → blank → CTA (question/invitation) → blank → up to 3 hashtags.
-**Announcement:** bold Unicode headline → what it is (2 sentences) → 🔹×3 key points → why it matters → CTA/link → 1-2 tags.
-**Regenerative / LIFEAI (Dave voice):** manifesto opener (one line) → blank → problem with the old way (2-3 lines) → blank → 🌱 what we're building / how it differs / who it serves → blank → the stakes (one line) → blank → direct personal invitation → blank → `𝗗𝗮𝘃𝗲 𝗟𝗮𝗱𝗼𝘂𝗰𝗲𝘂𝗿 | LIFEAI | 𝘓𝘪𝘧𝘦 𝘣𝘦𝘧𝘰𝘳𝘦 𝘗𝘳𝘰𝘧𝘪𝘵𝘴` → `#RegenerativeCapital #PlaceRegeneration #LIFEAI`.
-
-### Length
-Thought leadership 900–1300 chars · Announcement 400–700 · Hot take 200–400 · Long-form native 1200–2000 words.
-
-### Do NOT
-❌ `**bold**` (literal asterisks) · ❌ `# Heading` (literal hash) · ❌ `-`/`•` bullets (use emoticons) · ❌ LaTeX · ❌ >5 hashtags · ❌ tagging people unless asked · ❌ walls of text (max 3 lines/paragraph).
-
----
-
-## § Twitter / X
-
-### Core Principles
-- 280 characters hard limit per tweet. Hook is everything — first 5 words decide readership.
-- Threads are for depth; single tweets for punchy takes. No markdown renders — plain text only.
-- Minimal hashtags (1-2 max, only if genuinely searchable). End with a provocation/question when possible.
-
-### Character Counting
-- URLs always count as 23 chars (t.co). Media doesn't count. Line breaks = 1 char each. Leave a 10-15 char buffer — aim for 265 max.
-
-### Single Tweet Structure
-**Hot take:** provocative opener (≤10 words) → the turn/evidence (1-2 lines) → the landing.
-**Announcement:** bold claim first → why it matters (1 line) → link/CTA.
-**Regenerative (Dave):** manifesto line → old way vs new way → the invitation.
-
-### Thread Structure
-Tweet 1 = strongest hook (number `(1/7)` or use 🧵) → tweet 2 = context → tweets 3-n = one idea each → last tweet = recap + CTA/question. Each tweet must stand alone. Max 7-10 tweets — longer = blog post.
-
-### Hook Formulas
-Contrarian ("Most investors are wrong about regenerative land.") · Hard number ("We just unlocked $40M in stranded rural capital.") · Open loop ("Something broke in ESG investing. Here's what it missed:") · Bold declaration · Question · List tease ("3 things institutional capital gets wrong about place:").
-
-### Hashtags & Emoji
-Max 2 hashtags, end of tweet only, active communities only. Preferred LIFEAI tags: `#RegenerativeCapital #PlaceRegeneration #ImpactInvesting #RegenFinance`. Emoji 1-2 max as punctuation: 🌱 regen · 💰 capital · 🏗️ development · ⚡ urgency.
-
-### Do NOT
-❌ markdown · ❌ hashtag spam · ❌ "RT if you agree" · ❌ threads >10 tweets · ❌ passive openers ("I wanted to share…") · ❌ tagging without reason · ❌ ever exceed 280 chars (always count).
-
----
-
-## § Slack / Teams
-
-### Core Principles
-- Brevity is respect (people read on their phone). One message = one thought. Use threads. Emoji are punctuation, not decoration.
-
-### Slack Markdown (actually renders)
-Bold `*bold*` · Italic `_italic_` · Strikethrough `~strike~` · Inline code `` `code` `` · Code block ```` ``` ```` · Blockquote `> text` · Bullets `• item`.
-Note: `**bold**` (double asterisk) does NOT render in Slack — use single `*bold*`.
-
-### Length
-Quick question 1-2 lines · Status update 3-5 lines (bullets if 3+ items) · Decision request 5-8 lines · anything longer → thread or doc link.
-
-### Structure by Type
-**Question:** `Quick question on [topic] — [one-sentence question]?`
-**Status update:** `*[Project]* update:` then `• Done: …` `• In progress: …` `• Blocked on: … — need [thing] from [person]`.
-**Decision needed:** `Need a call on [topic] by [time].` → `Context: …` → `Options: A) … B) …` → `My lean: …` → `[@person] — your call.`
-**Async FYI:** `FYI: [what happened] — no action needed. [link]`
-
-### Emoji as Punctuation
-✅ done/approved · ❌ no/blocked · 🔁 in progress · 👀 reviewing · 🙋 I'll take it · 📌 important · ⚡ urgent · 🤔 needs discussion · 💬 let's talk · 🔗 link follows.
-
-### Do NOT
-❌ walls of text (use a doc) · ❌ `**double asterisk bold**` · ❌ formal greetings ("Hi team, I hope everyone…") · ❌ passive asks ("Would anyone be able to…" → "Can you [name]…") · ❌ emoji overload (max 2-3) · ❌ cross-posting the same message.
-
----
-
-## § Email — External
-
-### Core Principles
-- Plain prose, no markdown. Subject line determines open rate. First sentence justifies the email. One ask per email. Short paragraphs (2-4 sentences). Formal but human.
-
-### Structure
-`Subject:` → greeting (`Dear [First Name],`) → opening (1 sentence: context/connection) → body (¶1 reason, ¶2 value/evidence, ¶3 the one ask) → forward-looking closing line → sign-off, full name, title, org, 2-3 contact lines.
-
-### Subject Lines
-6-10 words; personalize. Cold: `[hook] — [value]` ("Regenerative capital for [Property]"). Follow-up: `Following up — [topic]`. Intro: `Introduction: [name] — [relevance]`. Request: `[Action] request — [context]`. Never use "Quick question" / "Checking in" / "Hope you're well" as a subject; never all caps.
-
-### Tone by Recipient
-Institutional/Family Office → formal, precise, lead with thesis + data · Foundation → warm, lead with impact · Government → neutral, process-oriented · Legal → exact, every claim qualified · Peer → professional, warmer · Media → newsworthy first.
-
-### Openers (use/adapt)
-"I'm reaching out because [specific reason tied to them]." · "We met at [event] — following up on [topic]." · "[Mutual contact] suggested I reach out regarding [topic]." Never open with "I hope this email finds you well." / "My name is X and I work at Y." / "I wanted to reach out to…".
-
-### Closings & Sign-offs
-"I'd welcome a 20-minute call at your convenience." / "Would you be open to a brief call this week or next?" Sign-offs: Respectfully / Yours sincerely (most formal) · Kind regards / Best regards (standard) · Warm regards (warm) · "In service of place," / "Regeneratively," (LIFEAI/Dave).
-LIFEAI signature: `Dave Ladouceur` / `Founder, LIFEAI | Place Regeneration Trust` / `dave@life.ai | lifeai.com` / `Life before Profits.`
-
-### Length
-Cold outreach 150–250 words · Follow-up 80–150 · Proposal intro 250–400 · Investor update 300–500.
-
-### Do NOT
-❌ markdown · ❌ walls of text · ❌ >1 ask · ❌ passive voice on the ask · ❌ sycophantic openers/closers · ❌ unspecified attachments.
-
----
-
-## § Email — Internal
-
-### Core Principles
-- Short, direct, action-oriented. Bullets fine and often preferred. Scannable subject. One topic per email. No pleasantries unless warranted.
-
-### Structure
-`Subject: [Action word] — [Topic]` (e.g. "Decision needed — Supabase migration") → optional 1-line context → direct body (2-4 short paragraphs or bullets) → clear next step on its own line → casual sign-off (Thanks / Dave / —D).
-
-### Subject Prefixes (use consistently)
-`ACTION:` recipient must do something · `DECISION:` approval/choice needed · `FYI:` informational · `URGENT:` same-day · `QUESTION:` single question · `UPDATE:` status. Example: `ACTION: Review PRT NAV calc before Friday`.
-
-### Bullets & Tone
-One idea per bullet, max 1 line, lead with the verb (Review… / Confirm… / Send…), no sub-bullets unless necessary. Direct without blunt; no "per my last email" energy; emoji sparingly.
-
-### Length
-Quick update 3-6 lines · Task assignment 6-12 · Internal brief ≤12-20 (longer → a doc).
-
-### Do NOT
-❌ markdown · ❌ long preambles · ❌ CC-ing everyone · ❌ "as per my previous email" · ❌ formal sign-offs internally.
-
----
-
-## § Pitch Deck
-
-### Core Principles
-- One slide, one job. Headline IS the message (a statement, not a label). 3 bullets max. No prose on slides (sentences → speaker notes). Visual-first language. Investor slides lead with thesis, support with evidence.
-
-### Slide Anatomy
-Headline (full sentence takeaway, 12-15 words) → visual/data area (chart/diagram/table) → ≤3 bullets (fragments, not sentences) → source/footnote if cited (small, bottom-right).
-
-### Headlines Are Assertions, Not Labels
-❌ "Market Opportunity" → ✅ "A $4T regenerative land market with no institutional gateway" · ❌ "Our Solution" → ✅ "PRT converts stranded land assets into regenerative yield" · ❌ "The Team" → ✅ "30 years of place-based development across 3 continents" · ❌ "Financial Returns" → ✅ "8–12% cash yield with embedded ecological appreciation".
-
-### Standard Deck Architecture (PRT/LIFEAI)
-1 Cover · 2 The Problem · 3 The Opportunity (market + gap) · 4 Our Solution · 5 How It Works (process diagram) · 6 Traction/Proof · 7 Market/TAM · 8 Business Model · 9 Financial Projections · 10 Capital Structure (A/B/C sleeves) · 11 The Team · 12 The Ask (amount, use of funds, timeline) · 13 Appendix.
-
-### Visual & Bullet Style
-Describe visuals explicitly, e.g. `[VISUAL: horizontal capital-stack diagram — Lane A Senior Secured $Xm 6% DAF | Lane B Mezz $Xm 9% Family Office | Lane C Equity $Xm carry GP]`. Bullets: fragments, no periods, lead with metric ("8–12% cash yield") or verb ("Converts stranded assets…"), max 8 words.
-Speaker notes: full conversational sentences expanding the slide — `[SPEAKER NOTE: …]`.
-Typography (for JSX/design): Headline 28-36pt bold · Bullets 18-22pt · Data labels 14-16pt accent · Source 10-11pt gray · Section label 11-12pt caps.
-
-### Do NOT
-❌ prose on slides · ❌ >3 bullets · ❌ label headlines · ❌ >2 font sizes in bullet area · ❌ generic stock-photo descriptions · ❌ animations unless requested.
-
----
-
-## § Word / DOCX
-
-> **This section = how to STRUCTURE Word content (Styles, hierarchy, tables).** To
-> generate an actual `.docx` file, or convert `.docx ↔ .md`, hand off to **`rdc:convert`
-> (build-corpus)** — channel-formatter does not produce or parse binary files.
-
-### Core Principles
-- All formatting via **Word Styles** (never raw bold alone). Heading hierarchy is semantic (Heading 1/2/3, not font size). Tables use Word Table Styles, not markdown. Equations via Equation Editor (OMML). Explicit page structure (section breaks, headers/footers, page numbers).
-
-### Style Hierarchy
-`Title` (one, top) · `Subtitle` · `Heading 1/2/3/4` · `Normal` / `Body Text` · `List Bullet` / `List Bullet 2` · `List Number` / `List Number 2` · `Quote`/`Block Text` · `Caption` · `Table Grid` · `Intense Quote` (callouts) · `Header`/`Footer`.
-
-### Document Types
-**Executive Report/White Paper:** Title + Subtitle → page break → H1 Executive Summary → H1 TOC (auto field) → page break → numbered H1/H2 sections with `Table Grid` tables + captions → H1 References (APA) → H1 Appendices.
-**Memo:** Title "MEMORANDUM" → bold-label TO/FROM/DATE/RE → rule → H2 Purpose / Background / Recommendation (List Bullet) → sign-off.
-**Investment/Deal Doc (PRT/RDC):** Title + Subtitle "Confidential | Prepared by LIFEAI | Date" → page break → H1 Executive Summary → H1 Investment Thesis (H2 Market Opportunity, H2 Regenerative Framework + Five Capitals `Table Grid`) → H1 Financial Structure (H2 Capital Stack tranche table, H2 NAV Architecture w/ Equation objects) → H1 Risk Framework (risk matrix) → H1 Governance → H1 Appendices.
-**Letter (formal):** sender block → date → recipient block → `Dear [Name],` → opening (purpose) → body → closing (next steps) → `Sincerely,` → signature.
-
-### Tables, Equations, Page Setup
-Tables: `Table Grid`, bold shaded header (`#F2F2F2`), left-align text / right-align numbers, `Caption` below ("Table X: …"). Equations: specify as Equation Editor objects, e.g. `[EQUATION OBJECT: αPRT = (regen_yield − WACC_conventional) + Σ(stewardship_delta)]` with caption. Page setup: Letter/A4, 1" margins, Calibri 11 / Times 12 body, 1.15–1.5 spacing, header (title left, page# right), footer (org left, date right), numbering from page 2.
-
-### Do NOT
-❌ manual bold instead of a Style for headings · ❌ tabs/spaces for indentation · ❌ `Shift+Enter` between paragraphs (use spacing) · ❌ markdown tables · ❌ inline LaTeX · ❌ >3 heading levels without justification.
-
----
-
-## § PDF Report
-
-### Core Principles
-- Final, fixed, reader-facing. Structure non-negotiable: executive summary → body → references. All sections numbered; all tables/figures captioned. Formal prose (no bullets as primary structure). Page numbers, headers, footers required.
-
-### Structure
-Cover (title, type, org+logo, date | version | confidentiality) → Executive Summary (≤1 page: what it is, 3-5 findings, recommendations) → TOC (auto, clickable) → numbered body (1 Introduction/Context · 2-4 core · 5 Conclusions) → References (APA 7th, numbered) → Appendices (A, B, C…).
-Section numbering: hierarchical decimal, max 3 levels (1.1.1), never 4.
-
-### Tables, Figures, Equations
-Tables: title above (`Table X: …`), bold shaded header, `Source:` below if external, `Note:` if needed. Figures: caption below (`Figure X: …`), source, bracketed alt-text. Equations: labeled `(Equation 1)` with a `Where:` variable key.
-Typography: Georgia 11 / Garamond 12 body; H1 18 bold, H2 14 bold, H3 12 bold italic; 1.5 body spacing; 1.25" binding margin; page numbers bottom-center from page 2; footer `Org | Confidential | Date`.
-Confidentiality labels: `CONFIDENTIAL` · `CONFIDENTIAL — For Recipient Only` · `FOR DISCUSSION PURPOSES ONLY` · `NOT FOR DISTRIBUTION` · (none = public).
-
-### Do NOT
-❌ markdown in final content · ❌ unnumbered sections · ❌ tables without titles/sources · ❌ figures without captions · ❌ invented data (mark "Illustrative") · ❌ exec summaries over one page · ❌ incomplete citations.
-
----
-
-## § Web / Landing Page
-
-### Core Principles
-- Scannable first, readable second. Above the fold is everything. One page = one conversion goal. SEO-aware without keyword stuffing. CTA appears at least twice (above fold + bottom). Mobile-first (short paragraphs/sentences).
-
-### Page Structure
-Hero (H1 value prop 6-10 words · subheadline 1-2 sentences · CTA button verb+object · optional hero media) → Social proof/trust bar → Problem (name the pain, 2-3 short paragraphs or 3 cards) → Solution (the shift; icon + headline + 1-2 sentences feature/benefit pairs) → How It Works (3-4 numbered steps, 1 sentence each) → Proof/Results (large number + short label: "340 acres restored") → CTA section (urgency headline + friction-reducer subtext + button) → Footer.
-
-### Headlines & Body
-H1 ≤8 words (benefit, not feature, active voice, power words: transform/unlock/regenerate/built/proven) · H2 ≤12 · H3 ≤6. Body: paragraphs ≤3 sentences/50 words, sentences ≤20 words, second person ("you/your"), grade 8-10 reading level.
-LIFEAI/PRT examples: H1 "The Capital Stack for Regenerative Land" · H2 "How PRT Turns Stranded Assets Into Living Investments" · H3 "Place Readiness" / "Covenant-Protected Yield" / "Verified Impact".
-
-### CTA & SEO
-CTA copy: verb + specific object ("Book Your 20-Minute Call", "See the Capital Structure", "Download the Investment Brief") — never "Submit"/"Click here"/"Learn more". SEO: primary keyword in H1 + first paragraph, secondary in H2s/body naturally, meta title 50-60 chars, meta description 150-160 chars, descriptive alt text on all images.
-
-### Do NOT
-❌ paragraphs >3 sentences above the fold · ❌ passive-voice CTAs · ❌ jargon without translation · ❌ carousels/sliders above the fold · ❌ >2 CTAs per section · ❌ generic headlines ("Welcome to Our Website") · ❌ markdown in final web copy (deliver clean prose with labeled sections).
-
----
-
-## Provenance
-
-Ported into rdc-skills from the `output-channel-formatter` claude.ai skill (frontend-design-skills pack).
-Originally a multi-file skill (SKILL.md + `references/*.md`); inlined into this single self-contained
-SKILL.md so it works identically across the CLI plugin, the claude.ai web client (via `rdc_skill_get`),
-and Codex — the rdc-skills MCP serves only SKILL.md, so all channel rules must live here.
+---
+name: rdc:channel-formatter
+description: "Usage `rdc:channel-formatter <channel|pack> [content]` — Apply precise, channel-native formatting and content repurposing to any output: LinkedIn, Twitter/X, Slack/Teams, Email (external/internal), Pitch Deck slides, Word/DOCX, PDF Report, Web/Landing Page, and multi-output content packs. Use EVERY TIME the user names an output channel/platform/document type, asks to 'write a post', 'draft an email', 'format this for', 'make this a slide', 'send to LinkedIn', 'write a tweet', 'social media post', 'turn this article into posts', 'make a content pack', 'make a social pack', or 'repurpose this article'. Each channel has its own Unicode strategy, emphasis system, length limits, and structure — never apply generic markdown to channel-specific output. This skill FORMATS, STRUCTURES, and REPURPOSES text only — for actual .docx/.pptx ↔ Markdown FILE conversion (either direction) use `rdc:convert` (build-corpus), not this skill. Self-contained: all channel and pack rules are inlined below (no external reference files)."
+---
+
+> **⚠️ OUTPUT CONTRACT (READ FIRST):** `guides/output-contract.md`
+> The formatted deliverable IS the output. Produce the channel-correct text directly.
+> No tool-call narration, no raw markdown wrappers around channel output, no log dumps.
+
+> If dispatching subagents or running as a subagent: read `{PROJECT_ROOT}/.rdc/guides/agent-bootstrap.md` first.
+
+> **Sandbox contract:** This skill honors `RDC_TEST=1` per `guides/agent-bootstrap.md` § RDC_TEST Sandbox Contract. It writes no files and makes no external calls; under `RDC_TEST=1` behavior is unchanged (pure text transform).
+
+# Channel Formatter
+
+Format or repurpose output precisely for the target channel. This skill is
+**self-contained** — every channel and pack rule is inlined below. Detect the
+channel or pack, jump to its section, apply its rules exactly. Never apply
+generic markdown to a channel that doesn't render it.
+
+---
+
+## When to Use
+
+Use this skill whenever the user asks to write, format, adapt, or repurpose
+content for a specific channel, platform, document type, or output pack.
+
+Common triggers:
+- "write a LinkedIn post"
+- "turn this article into social posts"
+- "make a content pack"
+- "make a social pack"
+- "repurpose this article"
+- "write a tweet/thread"
+- "draft an email"
+- "format this for Slack"
+- "make this a slide"
+- "write landing-page copy"
+
+Do not use this skill for binary file conversion, PDF rendering, Brochurify
+orchestration, or brochure JSX authoring; delegate those to the specialist
+skills named in the scope boundary.
+
+## Arguments
+
+`rdc:channel-formatter <channel|pack> [content]`
+
+- `<channel|pack>`: optional target such as `linkedin`, `twitter`,
+  `twitter-thread`, `slack`, `email-ext`, `email-int`, `pitch-deck`, `word`,
+  `pdf-report`, `web`, `strict-format`, `social-pack`, `campaign-pack`,
+  `exec-pack`, or `launch-pack`.
+- `[content]`: optional source text, path reference, article, report, transcript,
+  rough draft, or surrounding conversation content.
+
+## Channel Detection
+
+| If user says / context is...                          | Channel          | Go to section            |
+|-------------------------------------------------------|------------------|--------------------------|
+| LinkedIn, post, share on LinkedIn                     | LinkedIn         | [§ LinkedIn](#-linkedin) |
+| Tweet, Twitter, X post, thread                        | Twitter/X        | [§ Twitter / X](#-twitter--x) |
+| Slack, Teams, Discord, chat message                   | Slack/Teams      | [§ Slack / Teams](#-slack--teams) |
+| Email to client/investor/external, cold outreach      | Email (External) | [§ Email — External](#-email--external) |
+| Email to team, internal, Slack alternative            | Email (Internal) | [§ Email — Internal](#-email--internal) |
+| Pitch deck, slide, investor slide, one-pager slide    | Pitch Deck       | [§ Pitch Deck](#-pitch-deck) |
+| Word, DOCX, document, report, memo, letter, contract  | Word/DOCX        | [§ Word / DOCX](#-word--docx) |
+| PDF report, annual report, white paper                | PDF Report       | [§ PDF Report](#-pdf-report) |
+| Website, landing page, web copy                       | Web/Landing Page | [§ Web / Landing Page](#-web--landing-page) |
+| Artifact, JSX, React component, Claude artifact       | Artifact/JSX     | → use the `jsx-author` / `impeccable` skill instead |
+
+### Pack Detection
+
+| If user says / context is...                          | Mode             | Go to section                 |
+|-------------------------------------------------------|------------------|-------------------------------|
+| social pack, content pack, posts from this, repurpose this article for social | `social-pack` | [§ Content Repurposing Packs](#-content-repurposing-packs) |
+| campaign pack, launch campaign, article into email + social + web | `campaign-pack` | [§ Content Repurposing Packs](#-content-repurposing-packs) |
+| exec pack, executive pack, leadership summary, internal briefing | `exec-pack` | [§ Content Repurposing Packs](#-content-repurposing-packs) |
+| launch pack, announce this, launch posts              | `launch-pack`    | [§ Content Repurposing Packs](#-content-repurposing-packs) |
+| strict format, preserve wording, do not rewrite       | `strict-format`  | [§ Repurposing Modes](#-repurposing-modes) |
+
+---
+
+## Workflow
+
+1. **Detect channel or pack mode** from the request using the tables above.
+2. **Classify the source**: already-drafted copy, long article/report, transcript, notes, or brief.
+3. **For thin or under-specified long sources**, enrich before writing instead
+   of inventing context:
+   - If the source names a corpus path, read it.
+   - Otherwise search/read the approved corpus first when available. Resolve
+     `CORPUS_ROOT` (usually `H:/My Drive/global-corpus`) and
+     `LOCAL_CORPUS_ROOT` (usually `C:/Dev/local-corpus`) before searching the
+     repo. If environment variables are not visible, try those default paths.
+   - The corpus search must use an explicit corpus path in the tool arguments
+     (`H:/My Drive/global-corpus` or `C:/Dev/local-corpus`). A relative repo
+     search does not count as enrichment.
+   - Use `Grep`/`Glob`/`Read` or the environment's equivalent tools so the
+     transcript records what was consulted.
+   - Use web search only when the requested output needs current/public facts,
+     the user asks for external context, or corpus context is absent and network
+     search is allowed. Cite/use only what the search actually supports.
+   - Searching only the project repo is not enough for enrichment unless the
+     source itself points to a repo-local corpus file.
+   - If neither corpus nor web context is available, keep the output sparse and
+     mark missing context in assumptions; do not fill gaps creatively.
+4. **For long sources**, extract thesis, audience, proof points, CTA, constraints, and factual risks before writing.
+5. **Jump to the target channel or pack section** below and apply all rules exactly — do not rely on memory.
+6. **Never mix** markdown conventions across channels.
+7. If the channel or pack is ambiguous, ask once: "Is this for [Channel A] or [Channel B]?"
+8. Produce the formatted or repurposed output directly as the deliverable.
+9. Treat extraction notes and source-fidelity guardrails as internal scratchwork.
+   Do not print a "Long-Source Extraction Checklist", factual guardrail list, or
+   absent-topic list unless the user explicitly asks for analysis.
+
+## Hard Rules (all channels)
+
+- Never apply generic markdown (`##`, `**`, `-`, etc.) to LinkedIn, Email, Slack, or Twitter output.
+- Never use raw LaTeX in LinkedIn, Email, Slack, or Twitter — use Unicode math symbols.
+- Never use Word Styles notation in plain-text channels.
+- Always match the tone register of the channel (formal ≠ casual ≠ punchy).
+- For LIFEAI/PRT/RDC content, maintain REGEN-MODE voice unless Author-Mode is active.
+- When repurposing a long source, preserve the source thesis and never invent unsupported statistics, quotes, names, dates, citations, results, or commitments.
+- If the source lacks a needed CTA, audience, proof point, or date, either use a clearly generic placeholder or state the assumption before the polished output.
+
+> ## ⛔ Scope boundary — this skill FORMATS text; it does NOT convert files
+> This skill governs **how to structure and format content** for a channel. It never
+> reads or writes binary office files. For the actual **Word ↔ Markdown (and .pptx/.ppt)
+> file conversion — in either direction — use `build-corpus` / the `rdc:convert` skill**,
+> never this one. So: "convert this .docx to markdown" / "turn this markdown into a Word
+> doc" → **`rdc:convert`**. "format this content the Word way / write it for LinkedIn" →
+> this skill. The Word/DOCX and PDF sections below describe target structure only; producing
+> the actual `.docx`/`.pdf` artifact is `rdc:convert` / `rdc:brochure`, not channel-formatter.
+>
+> **Work in Markdown, materialize at the end.** Draft and format everything as Markdown
+> ("MD time" — cheap tokens, clean diffs); only hand off to `rdc:convert` to emit the
+> `.docx`/`.pptx` as the last step. build-corpus round-trips losslessly, so this costs
+> no fidelity. See the build-corpus agent primer (`AGENTS.md`).
+>
+> Specialist routing:
+> - Office/Markdown file conversion → `rdc:convert`
+> - HTML/folder/zip/URL to PDF brochure rendering → `rdc:brochure`
+> - Brochurify orchestration jobs → `rdc:brochurify`
+> - Brochure JSX using `@lifeai/brochure-kit` → `lifeai-brochure-author`
+
+---
+
+## § Repurposing Modes
+
+Use these modes when the input is longer than the requested output, such as an
+article, memo, transcript, report, rough draft, or notes.
+
+### `strict-format`
+Preserve the source wording and argument as much as the channel allows. Use this
+when the user says "format only", "do not rewrite", "preserve wording", or when
+legal/compliance precision matters. You may adjust line breaks, headings,
+emphasis, bullet symbols, and channel-specific structure, but do not add new
+claims or reframe the argument.
+
+### `single-channel`
+Repurpose the source into one target channel. Extract the usable argument,
+choose the strongest hook for that channel, compress or expand to the channel's
+native length, and produce one finished output.
+
+### `social-pack`
+Repurpose one source into a coordinated set of social outputs:
+- LinkedIn thought-leadership post
+- Short LinkedIn announcement or teaser variant
+- Twitter/X single post
+- Twitter/X thread
+- Slack/Teams internal share
+
+### `campaign-pack`
+Repurpose one source into a small campaign kit:
+- Everything in `social-pack`
+- External email intro/blurb
+- Web excerpt
+- SEO meta title and meta description
+- 3 CTA variants
+
+### `exec-pack`
+Repurpose one source for internal leadership alignment:
+- Internal email summary
+- Slack/Teams update
+- Executive-summary paragraph
+- 3 talking points
+- Decision/ask line if the source supports one
+
+### `launch-pack`
+Repurpose one source for an announcement:
+- LinkedIn launch post
+- Twitter/X launch post
+- Slack/Teams launch note
+- External email blurb
+- Web hero headline/subheadline/CTA
+- 3 CTA variants
+
+### Long-Source Extraction Checklist
+Before writing from a long source, identify internally:
+- **Thesis:** the central argument or announcement
+- **Audience:** who this is for
+- **Proof:** facts, examples, data, names, dates, or quotes explicitly present
+- **CTA:** what the reader should do next
+- **Tone:** formal, conversational, punchy, executive, regenerative, etc.
+- **Constraints:** length, compliance, brand voice, channel quirks
+- **Gaps:** missing proof, missing CTA, unsupported claims, unclear audience
+
+### Source-Fidelity Rules
+- Do not invent statistics, dates, quotes, citations, partnerships, revenue,
+  legal claims, customer names, or outcomes.
+- Do not infer adjacent finance/reporting concepts that sound plausible but are
+  absent from the source, such as pro forma assumptions, reporting cadence,
+  offset mechanisms, governance structures, verification status, or portfolio
+  implications.
+- If corpus or web context was consulted, use it only for facts it directly
+  supports. Do not blend generic domain knowledge into the source as if it were
+  documented.
+- If the source provides a theme label without an explanation, keep the theme
+  label or paraphrase it conservatively. Do not add explanatory clauses that
+  define how it works, who governs it, how it is verified, how often it reports,
+  or what financial model it opposes unless those details are explicit.
+- Do not contrast patient capital with quarters, earnings calls, fund cycles,
+  exits, venture speed, or portfolio management unless those words or concepts
+  appear in the source or consulted corpus/web material.
+- Do not upgrade tentative language into certainty.
+- Do not turn illustrative examples into facts.
+- Preserve caveats when they affect meaning.
+- If a stronger hook needs a proof point the source does not provide, write a
+  proof-neutral hook instead.
+- If the source says a topic is absent, excluded, or not mentioned, treat that
+  as an internal guardrail. Do not repeat the absent topic in public-facing
+  channel copy, assumptions, caveats, notes, headings, or summaries unless the
+  user explicitly asks for a contrast, compliance note, or risk disclosure.
+  Strip absent-topic lists from the deliverable entirely.
+- When assumptions are material, include a short "Assumptions:" line before the
+  deliverable rather than burying uncertainty in polished copy.
+- Do not output your extraction checklist. The deliverable should start with the
+  requested channel/pack output, except for a brief "Assumptions:" line when a
+  material gap affects the copy.
+- If you include an "Assumptions:" line, limit it to missing useful inputs such
+  as org name, audience, location, date, CTA, or link. Never list absent,
+  excluded, or not-mentioned topics in the assumptions line. If the only
+  uncertainty is an absent-topic guardrail, omit the assumptions line.
+
+---
+
+## § Content Repurposing Packs
+
+When a pack is requested, label each output clearly and make every item
+channel-native. A pack is not a generic summary repeated in several lengths.
+
+### `social-pack` Output Shape
+1. `LinkedIn thought-leadership post` — 900-1300 characters, hook-first, white space, Unicode emphasis only when useful, up to 3 hashtags.
+2. `LinkedIn short variant` — 400-700 characters, announcement or teaser style.
+3. `Twitter/X single post` — 265 characters or fewer unless the user asks otherwise.
+4. `Twitter/X thread` — 5-7 tweets, each self-contained and below 280 characters.
+5. `Slack/Teams internal share` — 3-6 lines, direct, with a clear FYI/action/decision framing.
+
+### `campaign-pack` Output Shape
+1. Full `social-pack`
+2. `External email intro` — subject line plus 80-150 word blurb with one ask
+3. `Web excerpt` — 80-120 words, scannable, CTA-ready
+4. `Meta title` — 50-60 characters
+5. `Meta description` — 150-160 characters
+6. `CTA variants` — 3 verb-led CTAs
+
+### `exec-pack` Output Shape
+1. `Internal email summary` — subject plus 6-12 lines
+2. `Slack/Teams update` — 3-5 lines
+3. `Executive summary paragraph` — 100-150 words
+4. `Talking points` — 3 concise bullets/fragments
+5. `Decision or ask` — one line, only if supported by source
+
+### `launch-pack` Output Shape
+1. `LinkedIn launch post` — announcement structure
+2. `Twitter/X launch post` — single post
+3. `Slack/Teams launch note` — internal update
+4. `External email blurb` — subject plus short body
+5. `Web hero` — headline, subheadline, CTA
+6. `CTA variants` — 3 options
+
+### Pack Quality Rules
+- Vary hooks by channel; do not repeat the same first sentence everywhere.
+- Keep the source thesis consistent across all outputs.
+- Adapt CTA strength to the channel: softer on thought leadership, direct in
+  email/web, concise in Slack.
+- Use channel-specific formatting rules from the sections below.
+- If source proof is weak, use curiosity and framing instead of inflated claims.
+- Do not include extraction notes, guardrail notes, or absent-topic caveats in
+  the pack. Those are reasoning aids, not channel outputs.
+
+---
+
+## § LinkedIn
+
+### Core Principles
+- No markdown whatsoever — LinkedIn renders it as raw symbols.
+- Unicode is the ONLY way to achieve bold, italic, and special emphasis.
+- Emoticons are structural tools, not decoration — use them as bullet replacements and section markers.
+- Hook in line 1 — LinkedIn truncates after ~2 lines before "see more".
+- White space is content — single-line breaks create rhythm and readability.
+
+### Unicode Character Sets
+**Bold (Mathematical Bold)** — key terms, company names, headlines, CTAs:
+```
+𝗔𝗕𝗖𝗗𝗘𝗙𝗚𝗛𝗜𝗝𝗞𝗟𝗠𝗡𝗢𝗣𝗤𝗥𝗦𝗧𝗨𝗩𝗪𝗫𝗬𝗭
+𝗮𝗯𝗰𝗱𝗲𝗳𝗴𝗵𝗶𝗷𝗸𝗹𝗺𝗻𝗼𝗽𝗾𝗿𝘀𝘁𝘂𝘃𝘄𝘅𝘆𝘇
+𝟬𝟭𝟮𝟯𝟰𝟱𝟲𝟳𝟴𝟵
+```
+**Italic (Mathematical Italic)** — emphasis, quotes, sub-themes:
+```
+𝘈𝘉𝘊𝘋𝘌𝘍𝘎𝘏𝘐𝘑𝘒𝘓𝘔𝘕𝘖𝘗𝘘𝘙𝘚𝘛𝘜𝘝𝘞𝘟𝘠𝘡
+𝘢𝘣𝘤𝘥𝘦𝘧𝘨𝘩𝘪𝘫𝘬𝘭𝘮𝘯𝘰𝘱𝘲𝘳𝘴𝘵𝘶𝘷𝘸𝘹𝘺𝘻
+```
+**Bold Italic:**
+```
+𝘼𝘽𝘾𝘿𝙀𝙁𝙂𝙃𝙄𝙅𝙆𝙇𝙈𝙉𝙊𝙋𝙌𝙍𝙎𝙏𝙐𝙑𝙒𝙓𝙔𝙕
+𝙖𝙗𝙘𝙙𝙚𝙛𝙜𝙝𝙞𝙟𝙠𝙡𝙢𝙣𝙤𝙥𝙦𝙧𝙨𝙩𝙪𝙫𝙬𝙭𝙮𝙯
+```
+**Math / Equations** — Unicode symbols, never LaTeX: `× ÷ ± ≠ ≤ ≥ ≈ ∴ ∵ Σ Δ α β ∞ √ ² CO₂ H₂O → ← ⟺ % °`
+Example: instead of `E = mc^2` write `E = mc²`; instead of `\alpha_{PRT}` write `αPRT`.
+
+### Emoticons as Structure
+Section openers / bullets: 🔹 primary · 🔸 secondary · ▸ tertiary · ✅ result · ❌ contrast · ⚡ urgency · 🌱 regenerative (LIFEAI/PRT) · 💡 insight · 📌 takeaway · 🔑 key · 📊 data · 🏗️ development · 🤝 partnership · 💰 capital · 🌍 planetary.
+Dividers: `——————————————` · `▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬` · `· · · · · · · · · · · · · ·`
+
+### Post Structure
+**Standard (thought leadership):** hook line (no period, open loop) → blank → 3-6 short paragraphs (1-3 lines), emoticons as bullets → blank → insight/contrast line → blank → CTA (question/invitation) → blank → up to 3 hashtags.
+**Announcement:** bold Unicode headline → what it is (2 sentences) → 🔹×3 key points → why it matters → CTA/link → 1-2 tags.
+**Regenerative / LIFEAI (Dave voice):** manifesto opener (one line) → blank → problem with the old way (2-3 lines) → blank → 🌱 what we're building / how it differs / who it serves → blank → the stakes (one line) → blank → direct personal invitation → blank → `𝗗𝗮𝘃𝗲 𝗟𝗮𝗱𝗼𝘂𝗰𝗲𝘂𝗿 | LIFEAI | 𝘓𝘪𝘧𝘦 𝘣𝘦𝘧𝘰𝘳𝘦 𝘗𝘳𝘰𝘧𝘪𝘵𝘴` → `#RegenerativeCapital #PlaceRegeneration #LIFEAI`.
+
+### Length
+Thought leadership 900–1300 chars · Announcement 400–700 · Hot take 200–400 · Long-form native 1200–2000 words.
+
+### Do NOT
+❌ `**bold**` (literal asterisks) · ❌ `# Heading` (literal hash) · ❌ `-`/`•` bullets (use emoticons) · ❌ LaTeX · ❌ >5 hashtags · ❌ tagging people unless asked · ❌ walls of text (max 3 lines/paragraph).
+
+---
+
+## § Twitter / X
+
+### Core Principles
+- 280 characters hard limit per tweet. Hook is everything — first 5 words decide readership.
+- Threads are for depth; single tweets for punchy takes. No markdown renders — plain text only.
+- Minimal hashtags (1-2 max, only if genuinely searchable). End with a provocation/question when possible.
+
+### Character Counting
+- URLs always count as 23 chars (t.co). Media doesn't count. Line breaks = 1 char each. Leave a 10-15 char buffer — aim for 265 max.
+
+### Single Tweet Structure
+**Hot take:** provocative opener (≤10 words) → the turn/evidence (1-2 lines) → the landing.
+**Announcement:** bold claim first → why it matters (1 line) → link/CTA.
+**Regenerative (Dave):** manifesto line → old way vs new way → the invitation.
+
+### Thread Structure
+Tweet 1 = strongest hook (number `(1/7)` or use 🧵) → tweet 2 = context → tweets 3-n = one idea each → last tweet = recap + CTA/question. Each tweet must stand alone. Max 7-10 tweets — longer = blog post.
+
+### Hook Formulas
+Contrarian ("Most investors are wrong about regenerative land.") · Hard number ("We just unlocked $40M in stranded rural capital.") · Open loop ("Something broke in ESG investing. Here's what it missed:") · Bold declaration · Question · List tease ("3 things institutional capital gets wrong about place:").
+
+### Hashtags & Emoji
+Max 2 hashtags, end of tweet only, active communities only. Preferred LIFEAI tags: `#RegenerativeCapital #PlaceRegeneration #ImpactInvesting #RegenFinance`. Emoji 1-2 max as punctuation: 🌱 regen · 💰 capital · 🏗️ development · ⚡ urgency.
+
+### Do NOT
+❌ markdown · ❌ hashtag spam · ❌ "RT if you agree" · ❌ threads >10 tweets · ❌ passive openers ("I wanted to share…") · ❌ tagging without reason · ❌ ever exceed 280 chars (always count).
+
+---
+
+## § Slack / Teams
+
+### Core Principles
+- Brevity is respect (people read on their phone). One message = one thought. Use threads. Emoji are punctuation, not decoration.
+
+### Slack Markdown (actually renders)
+Bold `*bold*` · Italic `_italic_` · Strikethrough `~strike~` · Inline code `` `code` `` · Code block ```` ``` ```` · Blockquote `> text` · Bullets `• item`.
+Note: `**bold**` (double asterisk) does NOT render in Slack — use single `*bold*`.
+
+### Length
+Quick question 1-2 lines · Status update 3-5 lines (bullets if 3+ items) · Decision request 5-8 lines · anything longer → thread or doc link.
+
+### Structure by Type
+**Question:** `Quick question on [topic] — [one-sentence question]?`
+**Status update:** `*[Project]* update:` then `• Done: …` `• In progress: …` `• Blocked on: … — need [thing] from [person]`.
+**Decision needed:** `Need a call on [topic] by [time].` → `Context: …` → `Options: A) … B) …` → `My lean: …` → `[@person] — your call.`
+**Async FYI:** `FYI: [what happened] — no action needed. [link]`
+
+### Emoji as Punctuation
+✅ done/approved · ❌ no/blocked · 🔁 in progress · 👀 reviewing · 🙋 I'll take it · 📌 important · ⚡ urgent · 🤔 needs discussion · 💬 let's talk · 🔗 link follows.
+
+### Do NOT
+❌ walls of text (use a doc) · ❌ `**double asterisk bold**` · ❌ formal greetings ("Hi team, I hope everyone…") · ❌ passive asks ("Would anyone be able to…" → "Can you [name]…") · ❌ emoji overload (max 2-3) · ❌ cross-posting the same message.
+
+---
+
+## § Email — External
+
+### Core Principles
+- Plain prose, no markdown. Subject line determines open rate. First sentence justifies the email. One ask per email. Short paragraphs (2-4 sentences). Formal but human.
+
+### Structure
+`Subject:` → greeting (`Dear [First Name],`) → opening (1 sentence: context/connection) → body (¶1 reason, ¶2 value/evidence, ¶3 the one ask) → forward-looking closing line → sign-off, full name, title, org, 2-3 contact lines.
+
+### Subject Lines
+6-10 words; personalize. Cold: `[hook] — [value]` ("Regenerative capital for [Property]"). Follow-up: `Following up — [topic]`. Intro: `Introduction: [name] — [relevance]`. Request: `[Action] request — [context]`. Never use "Quick question" / "Checking in" / "Hope you're well" as a subject; never all caps.
+
+### Tone by Recipient
+Institutional/Family Office → formal, precise, lead with thesis + data · Foundation → warm, lead with impact · Government → neutral, process-oriented · Legal → exact, every claim qualified · Peer → professional, warmer · Media → newsworthy first.
+
+### Openers (use/adapt)
+"I'm reaching out because [specific reason tied to them]." · "We met at [event] — following up on [topic]." · "[Mutual contact] suggested I reach out regarding [topic]." Never open with "I hope this email finds you well." / "My name is X and I work at Y." / "I wanted to reach out to…".
+
+### Closings & Sign-offs
+"I'd welcome a 20-minute call at your convenience." / "Would you be open to a brief call this week or next?" Sign-offs: Respectfully / Yours sincerely (most formal) · Kind regards / Best regards (standard) · Warm regards (warm) · "In service of place," / "Regeneratively," (LIFEAI/Dave).
+LIFEAI signature: `Dave Ladouceur` / `Founder, LIFEAI | Place Regeneration Trust` / `dave@life.ai | lifeai.com` / `Life before Profits.`
+
+### Length
+Cold outreach 150–250 words · Follow-up 80–150 · Proposal intro 250–400 · Investor update 300–500.
+
+### Do NOT
+❌ markdown · ❌ walls of text · ❌ >1 ask · ❌ passive voice on the ask · ❌ sycophantic openers/closers · ❌ unspecified attachments.
+
+---
+
+## § Email — Internal
+
+### Core Principles
+- Short, direct, action-oriented. Bullets fine and often preferred. Scannable subject. One topic per email. No pleasantries unless warranted.
+
+### Structure
+`Subject: [Action word] — [Topic]` (e.g. "Decision needed — Supabase migration") → optional 1-line context → direct body (2-4 short paragraphs or bullets) → clear next step on its own line → casual sign-off (Thanks / Dave / —D).
+
+### Subject Prefixes (use consistently)
+`ACTION:` recipient must do something · `DECISION:` approval/choice needed · `FYI:` informational · `URGENT:` same-day · `QUESTION:` single question · `UPDATE:` status. Example: `ACTION: Review PRT NAV calc before Friday`.
+
+### Bullets & Tone
+One idea per bullet, max 1 line, lead with the verb (Review… / Confirm… / Send…), no sub-bullets unless necessary. Direct without blunt; no "per my last email" energy; emoji sparingly.
+
+### Length
+Quick update 3-6 lines · Task assignment 6-12 · Internal brief ≤12-20 (longer → a doc).
+
+### Do NOT
+❌ markdown · ❌ long preambles · ❌ CC-ing everyone · ❌ "as per my previous email" · ❌ formal sign-offs internally.
+
+---
+
+## § Pitch Deck
+
+### Core Principles
+- One slide, one job. Headline IS the message (a statement, not a label). 3 bullets max. No prose on slides (sentences → speaker notes). Visual-first language. Investor slides lead with thesis, support with evidence.
+
+### Slide Anatomy
+Headline (full sentence takeaway, 12-15 words) → visual/data area (chart/diagram/table) → ≤3 bullets (fragments, not sentences) → source/footnote if cited (small, bottom-right).
+
+### Headlines Are Assertions, Not Labels
+❌ "Market Opportunity" → ✅ "A $4T regenerative land market with no institutional gateway" · ❌ "Our Solution" → ✅ "PRT converts stranded land assets into regenerative yield" · ❌ "The Team" → ✅ "30 years of place-based development across 3 continents" · ❌ "Financial Returns" → ✅ "8–12% cash yield with embedded ecological appreciation".
+
+### Standard Deck Architecture (PRT/LIFEAI)
+1 Cover · 2 The Problem · 3 The Opportunity (market + gap) · 4 Our Solution · 5 How It Works (process diagram) · 6 Traction/Proof · 7 Market/TAM · 8 Business Model · 9 Financial Projections · 10 Capital Structure (A/B/C sleeves) · 11 The Team · 12 The Ask (amount, use of funds, timeline) · 13 Appendix.
+
+### Visual & Bullet Style
+Describe visuals explicitly, e.g. `[VISUAL: horizontal capital-stack diagram — Lane A Senior Secured $Xm 6% DAF | Lane B Mezz $Xm 9% Family Office | Lane C Equity $Xm carry GP]`. Bullets: fragments, no periods, lead with metric ("8–12% cash yield") or verb ("Converts stranded assets…"), max 8 words.
+Speaker notes: full conversational sentences expanding the slide — `[SPEAKER NOTE: …]`.
+Typography (for JSX/design): Headline 28-36pt bold · Bullets 18-22pt · Data labels 14-16pt accent · Source 10-11pt gray · Section label 11-12pt caps.
+
+### Do NOT
+❌ prose on slides · ❌ >3 bullets · ❌ label headlines · ❌ >2 font sizes in bullet area · ❌ generic stock-photo descriptions · ❌ animations unless requested.
+
+---
+
+## § Word / DOCX
+
+> **This section = how to STRUCTURE Word content (Styles, hierarchy, tables).** To
+> generate an actual `.docx` file, or convert `.docx ↔ .md`, hand off to **`rdc:convert`
+> (build-corpus)** — channel-formatter does not produce or parse binary files.
+
+### Core Principles
+- All formatting via **Word Styles** (never raw bold alone). Heading hierarchy is semantic (Heading 1/2/3, not font size). Tables use Word Table Styles, not markdown. Equations via Equation Editor (OMML). Explicit page structure (section breaks, headers/footers, page numbers).
+
+### Style Hierarchy
+`Title` (one, top) · `Subtitle` · `Heading 1/2/3/4` · `Normal` / `Body Text` · `List Bullet` / `List Bullet 2` · `List Number` / `List Number 2` · `Quote`/`Block Text` · `Caption` · `Table Grid` · `Intense Quote` (callouts) · `Header`/`Footer`.
+
+### Document Types
+**Executive Report/White Paper:** Title + Subtitle → page break → H1 Executive Summary → H1 TOC (auto field) → page break → numbered H1/H2 sections with `Table Grid` tables + captions → H1 References (APA) → H1 Appendices.
+**Memo:** Title "MEMORANDUM" → bold-label TO/FROM/DATE/RE → rule → H2 Purpose / Background / Recommendation (List Bullet) → sign-off.
+**Investment/Deal Doc (PRT/RDC):** Title + Subtitle "Confidential | Prepared by LIFEAI | Date" → page break → H1 Executive Summary → H1 Investment Thesis (H2 Market Opportunity, H2 Regenerative Framework + Five Capitals `Table Grid`) → H1 Financial Structure (H2 Capital Stack tranche table, H2 NAV Architecture w/ Equation objects) → H1 Risk Framework (risk matrix) → H1 Governance → H1 Appendices.
+**Letter (formal):** sender block → date → recipient block → `Dear [Name],` → opening (purpose) → body → closing (next steps) → `Sincerely,` → signature.
+
+### Tables, Equations, Page Setup
+Tables: `Table Grid`, bold shaded header (`#F2F2F2`), left-align text / right-align numbers, `Caption` below ("Table X: …"). Equations: specify as Equation Editor objects, e.g. `[EQUATION OBJECT: αPRT = (regen_yield − WACC_conventional) + Σ(stewardship_delta)]` with caption. Page setup: Letter/A4, 1" margins, Calibri 11 / Times 12 body, 1.15–1.5 spacing, header (title left, page# right), footer (org left, date right), numbering from page 2.
+
+### Do NOT
+❌ manual bold instead of a Style for headings · ❌ tabs/spaces for indentation · ❌ `Shift+Enter` between paragraphs (use spacing) · ❌ markdown tables · ❌ inline LaTeX · ❌ >3 heading levels without justification.
+
+---
+
+## § PDF Report
+
+### Core Principles
+- Final, fixed, reader-facing. Structure non-negotiable: executive summary → body → references. All sections numbered; all tables/figures captioned. Formal prose (no bullets as primary structure). Page numbers, headers, footers required.
+
+### Structure
+Cover (title, type, org+logo, date | version | confidentiality) → Executive Summary (≤1 page: what it is, 3-5 findings, recommendations) → TOC (auto, clickable) → numbered body (1 Introduction/Context · 2-4 core · 5 Conclusions) → References (APA 7th, numbered) → Appendices (A, B, C…).
+Section numbering: hierarchical decimal, max 3 levels (1.1.1), never 4.
+
+### Tables, Figures, Equations
+Tables: title above (`Table X: …`), bold shaded header, `Source:` below if external, `Note:` if needed. Figures: caption below (`Figure X: …`), source, bracketed alt-text. Equations: labeled `(Equation 1)` with a `Where:` variable key.
+Typography: Georgia 11 / Garamond 12 body; H1 18 bold, H2 14 bold, H3 12 bold italic; 1.5 body spacing; 1.25" binding margin; page numbers bottom-center from page 2; footer `Org | Confidential | Date`.
+Confidentiality labels: `CONFIDENTIAL` · `CONFIDENTIAL — For Recipient Only` · `FOR DISCUSSION PURPOSES ONLY` · `NOT FOR DISTRIBUTION` · (none = public).
+
+### Do NOT
+❌ markdown in final content · ❌ unnumbered sections · ❌ tables without titles/sources · ❌ figures without captions · ❌ invented data (mark "Illustrative") · ❌ exec summaries over one page · ❌ incomplete citations.
+
+---
+
+## § Web / Landing Page
+
+### Core Principles
+- Scannable first, readable second. Above the fold is everything. One page = one conversion goal. SEO-aware without keyword stuffing. CTA appears at least twice (above fold + bottom). Mobile-first (short paragraphs/sentences).
+
+### Page Structure
+Hero (H1 value prop 6-10 words · subheadline 1-2 sentences · CTA button verb+object · optional hero media) → Social proof/trust bar → Problem (name the pain, 2-3 short paragraphs or 3 cards) → Solution (the shift; icon + headline + 1-2 sentences feature/benefit pairs) → How It Works (3-4 numbered steps, 1 sentence each) → Proof/Results (large number + short label: "340 acres restored") → CTA section (urgency headline + friction-reducer subtext + button) → Footer.
+
+### Headlines & Body
+H1 ≤8 words (benefit, not feature, active voice, power words: transform/unlock/regenerate/built/proven) · H2 ≤12 · H3 ≤6. Body: paragraphs ≤3 sentences/50 words, sentences ≤20 words, second person ("you/your"), grade 8-10 reading level.
+LIFEAI/PRT examples: H1 "The Capital Stack for Regenerative Land" · H2 "How PRT Turns Stranded Assets Into Living Investments" · H3 "Place Readiness" / "Covenant-Protected Yield" / "Verified Impact".
+
+### CTA & SEO
+CTA copy: verb + specific object ("Book Your 20-Minute Call", "See the Capital Structure", "Download the Investment Brief") — never "Submit"/"Click here"/"Learn more". SEO: primary keyword in H1 + first paragraph, secondary in H2s/body naturally, meta title 50-60 chars, meta description 150-160 chars, descriptive alt text on all images.
+
+### Do NOT
+❌ paragraphs >3 sentences above the fold · ❌ passive-voice CTAs · ❌ jargon without translation · ❌ carousels/sliders above the fold · ❌ >2 CTAs per section · ❌ generic headlines ("Welcome to Our Website") · ❌ markdown in final web copy (deliver clean prose with labeled sections).
+
+---
+
+## Provenance
+
+Ported into rdc-skills from the `output-channel-formatter` claude.ai skill (frontend-design-skills pack).
+Originally a multi-file skill (SKILL.md + `references/*.md`); inlined into this single self-contained
+SKILL.md so it works identically across the CLI plugin, the claude.ai web client (via `rdc_skill_get`),
+and Codex — the rdc-skills MCP serves only SKILL.md, so all channel rules must live here.