Claude Skill

speak-as

Write a public post in the owner's own voice with a role model's style and idea palette layered on top, reading a style-palette file so the same engine works for any thinker the owner has profiled. Draft-first. Triggers: "/speak-as <name>", "write in the style of <X>".

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Download tonydzi-second-brain-starter-kit-skills_speak-as-7bc2efa.zip · 3 KB

Install

skills CLI npx skills add https://github.com/tonydzi/second-brain-starter-kit/tree/master/skills/speak-as
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install tonydzi-second-brain-starter-kit@llmmart
Git git clone https://github.com/tonydzi/second-brain-starter-kit.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole tonydzi/second-brain-starter-kit collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

/speak-as

🧒 When reporting to a non-technical owner: end with a child-simple "In plain words" recap (only in the reply TO him, NEVER inside the post draft).

A thin writing engine, parameterized by the role model: it takes the style palette of the chosen thinker + the owner's voice + his real material on the topic and writes ONE post as a mix — "frame/optics from the role model, voice/confessional tone/concreteness from the owner". One engine for any role model (one person today, anyone tomorrow). A manual command "write a post in the style of X about Y" (≠ the daily [[content-factory]], ≠ the /portret dossier).

[!warning] SAFETY FUSE (read this first)

  • ⭐ The owner's authorial voice is written by the top-tier model or stronger (a cheap model is forbidden for authorial text — a carve-out of model-routing-sonnet-grunt). Session weaker than that → delegate the WRITING itself to a top-model subagent (Agent, model:'opus'); the groundwork (resolving the role model, recall, choosing the platform) can run on the cheap model.
  • INFLUENCE, NOT copying. The role model's words = origin: external. Don't paste his phrases verbatim, don't pass his thoughts off as ours without rethinking them. The post stays the owner's post (origin: anton).
  • Preserve the owner — don't dissolve into the role model (his = the frame/optics, the owner's = the material and the voice).
  • Privacy from [[fb-diary-voice]] applies (names/amounts/sensitive material/live leads and deals must not be burned; abstract them up to an insight).
  • Draft-first — I hand over the draft, the owner posts it himself (publishing outbound = Tier-2).

Step 0 — Resolve the role model → the palette (cheap, no LLM)

  1. The role model's name → a latin slug (as in 07-People\person-<slug>.md).
  2. Look for the palette: $OBSIDIAN_VAULT/08-Templates/style-<slug>-influence.md.
    • Exists → load it (techniques + idea palette + "how to preserve the owner"). Carry on.
    • Missing → a STOP fork for the owner: "There is no style palette for
  3. Always load the owner's voice: [[fb-diary-voice]] — ⚠️ this is a MEMORY file on the system drive (~\.claude\projects\...\memory\fb-diary-voice.md), NOT a vault note; searching for it in the vault is pointless (the two-drives pitfall).

Step 1 — RECALL on the topic (the material MUST be the owner's, not generalities)

/ask "<topic>" (RAG) · /search "<words>" (exact words) · a fresh day from the [[content-factory]] inbox if it is a diary post. The groundwork can run on a cheap subagent; the text itself — top model.

Step 2 — Platform and length

Facebook diary ~3000–4500 chars (confessional, interwoven threads) · X — short (a hook formula + 1 thought) · Telegram content — medium. Not specified → default to the Facebook diary and say so.

Step 3 — Write the mix (top model)

From the role model (out of his palette file): the frame, the optics, the signature techniques, the idea boundaries. From the owner: the concrete detail of the day, hard self-irony, vivid images, confessional tone, 1–2 genuine questions at the end, honesty. Contested topics — handled with epistemic neutrality ([[epistemic-neutrality]]).

Step 4 — Self-check before handing it over (quality gate)

  • Does it sound like the owner, not like a cosplay of the role model? (role model outweighs him → rewrite in the owner's voice)
  • Is there a hook + a frame (formula/triad/theses) + 1–2 genuine questions?
  • Zero verbatim phrases from the role model; ideas rethought, not copied?
  • Privacy respected (no third-party names/amounts/live deals/sensitive material)?
  • Length fits the platform? No long dashes ([[no-long-dashes]]). Below the owner's bar → rewrite; don't hand over a weak draft.

Delivery

A clean post block (ready to copy) + one line: role model · platform · length · which technique was used. The owner posts/edits it himself. Optionally — 2–3 hook headline variants.

How to ADD a new role model (full cycle)

  1. /portret <Name> → the dossier person-<slug> + (synthesis) an insight-* note on "how he thinks".
  2. Synthesise the palette 08-Templates\style-<slug>-influence.md following the reference palette (style techniques + idea palette + "how to preserve the owner" + the influence-not-copy fuse). Back up before writing ([[vault-backup-rule]]).
  3. Done — /speak-as <Name> about <topic> works immediately (the same engine reads the new file).

Gates and links

  • Top model for the writing; groundwork on the cheap model. Draft-first; publication = Tier-2 (the owner himself).
  • The reference palette · the core-ideas insight note · the voice [[fb-diary-voice]] · the pipeline [[content-factory]] · the person dossier.

Like this skill? It is part of second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.

Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).

Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.

Files (second-brain-starter-kit)
  • SKILL.md 6.4 KB
    ---
    name: speak-as
    description: >-
      Write a public post in the owner's own voice with a role model's style and idea palette
      layered on top, reading a style-palette file so the same engine works for any thinker the
      owner has profiled. Draft-first. Triggers: "/speak-as <name>", "write in the style of <X>".
    license: MIT
    ---
    
    # /speak-as <role model> — a post in the owner's voice through the chosen thinker's optics
    
    > 🧒 **When reporting to a non-technical owner:** end with a child-simple "In plain words" recap (only in the reply TO him, NEVER inside the post draft).
    
    A thin writing engine, **parameterized by the role model**: it takes the **style palette of the chosen thinker** + **the owner's voice** + **his real material on the topic** and writes ONE post as a mix — "frame/optics from the role model, voice/confessional tone/concreteness from the owner". One engine for any role model (one person today, anyone tomorrow). A manual command "write a post in the style of X about Y" (≠ the daily [[content-factory]], ≠ the `/portret` dossier).
    
    > [!warning] SAFETY FUSE (read this first)
    > - ⭐ **The owner's authorial voice is written by the top-tier model or stronger** (a cheap model is forbidden for authorial text — a carve-out of `model-routing-sonnet-grunt`). Session weaker than that → delegate the WRITING itself to a top-model subagent (`Agent`, `model:'opus'`); the groundwork (resolving the role model, recall, choosing the platform) can run on the cheap model.
    > - **INFLUENCE, NOT copying.** The role model's words = `origin: external`. Don't paste his phrases verbatim, don't pass his thoughts off as ours without rethinking them. The post stays the owner's post (`origin: anton`).
    > - **Preserve the owner** — don't dissolve into the role model (his = the frame/optics, the owner's = the material and the voice).
    > - **Privacy** from [[fb-diary-voice]] applies (names/amounts/sensitive material/live leads and deals must not be burned; abstract them up to an insight).
    > - **Draft-first** — I hand over the draft, the owner posts it himself (publishing outbound = Tier-2).
    
    ## Step 0 — Resolve the role model → the palette (cheap, no LLM)
    1. The role model's name → a latin slug (as in `07-People\person-<slug>.md`).
    2. Look for the palette: `$OBSIDIAN_VAULT/08-Templates/style-<slug>-influence.md`.
       - **Exists** → load it (techniques + idea palette + "how to preserve the owner"). Carry on.
       - **Missing** → a STOP fork for the owner: "There is no style palette for <X> yet. Build one? → run `/portret <X>` (dossier + social media + DR), then I'll synthesise `style-<slug>-influence.md` and write in his style." Do not invent a style out of thin air.
    3. Always load the owner's voice: [[fb-diary-voice]] — ⚠️ this is a MEMORY file on the system drive (`~\.claude\projects\...\memory\fb-diary-voice.md`), NOT a vault note; searching for it in the vault is pointless (the two-drives pitfall).
    
    ## Step 1 — RECALL on the topic (the material MUST be the owner's, not generalities)
    `/ask "<topic>"` (RAG) · `/search "<words>"` (exact words) · a fresh day from the [[content-factory]] inbox if it is a diary post. The groundwork can run on a cheap subagent; the text itself — top model.
    
    ## Step 2 — Platform and length
    Facebook diary ~3000–4500 chars (confessional, interwoven threads) · X — short (a hook formula + 1 thought) · Telegram content — medium. Not specified → default to the Facebook diary and say so.
    
    ## Step 3 — Write the mix (top model)
    From the **role model** (out of his palette file): the frame, the optics, the signature techniques, the idea boundaries. From the **owner**: the concrete detail of the day, hard self-irony, vivid images, confessional tone, 1–2 genuine questions at the end, honesty. Contested topics — handled with epistemic neutrality ([[epistemic-neutrality]]).
    
    ## Step 4 — Self-check before handing it over (quality gate)
    - [ ] Does it sound like **the owner**, not like a cosplay of the role model? (role model outweighs him → rewrite in the owner's voice)
    - [ ] Is there a hook + a frame (formula/triad/theses) + 1–2 genuine questions?
    - [ ] **Zero verbatim phrases** from the role model; ideas rethought, not copied?
    - [ ] Privacy respected (no third-party names/amounts/live deals/sensitive material)?
    - [ ] Length fits the platform? No long dashes ([[no-long-dashes]]).
    Below the owner's bar → rewrite; don't hand over a weak draft.
    
    ## Delivery
    A clean post block (ready to copy) + one line: role model · platform · length · which technique was used. The owner posts/edits it himself. Optionally — 2–3 hook headline variants.
    
    ## How to ADD a new role model (full cycle)
    1. `/portret <Name>` → the dossier `person-<slug>` + (synthesis) an `insight-*` note on "how he thinks".
    2. Synthesise the palette `08-Templates\style-<slug>-influence.md` following the reference palette (style techniques + idea palette + "how to preserve the owner" + the influence-not-copy fuse). Back up before writing ([[vault-backup-rule]]).
    3. Done — `/speak-as <Name> about <topic>` works immediately (the same engine reads the new file).
    
    ## Gates and links
    - Top model for the writing; groundwork on the cheap model. Draft-first; publication = Tier-2 (the owner himself).
    - The reference palette · the core-ideas insight note · the voice [[fb-diary-voice]] · the pipeline [[content-factory]] · the person dossier.
    
    <!--kit-footer-->
    
    ---
    
    **Like this skill?** It is part of [second-brain-starter-kit](https://github.com/tonydzi/second-brain-starter-kit): the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with `npx skills add tonydzi/second-brain-starter-kit`. Everything is open source and free, so take what you need.
    
    Flagships worth a look on their own: [secondop-panel](https://github.com/tonydzi/secondop-panel) (a second opinion from a panel of external models), [claude-memory-tidy](https://github.com/tonydzi/claude-memory-tidy) (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tonydzi/telegram-mcp-kit) (your own Telegram over MCP in about 15 minutes).
    
    Author: **Anton Dziatkovskii**, Palo Alto AI Research Lab. Telegram [@tonydzi](https://t.me/tonydzi) - WhatsApp [+1 341 222 9178](https://wa.me/13412229178) - X [@Tony_Stef_](https://x.com/Tony_Stef_)
    
    **Engineers: want to test-drive this setup?** Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.