Skip to content
Back to search

ghostwriter

MarketplaceWorks with Claude CodeGitHub

Ghostwriter — a Claude Code plugin that keeps evidence-backed negotiation profiles of the people you meet, built from meeting transcripts and stored in your Notion. Never sends, every claim backed by a quote.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/rlfszn--ghostwriter/badge.svg)](/listings/rlfszn--ghostwriter)

HTML

<a href="/listings/rlfszn--ghostwriter"><img src="/api/listings/rlfszn--ghostwriter/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

As a plugin:

/plugin marketplace add rlfszn/ghostwriter
/plugin install ghostwriter@ghostwriter

Or copy the skills by hand (keep all seven folders together: they reference each other as siblings):

git clone https://github.com/rlfszn/ghostwriter
cp -r ghostwriter/skills/* ~/.claude/skills/

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 23 lines
  1. {
  2. "name": "ghostwriter",
  3. "owner": {
  4. "name": "Elia Rolfi"
  5. },
  6. "metadata": {
  7. "description": "Ghostwriter: evidence-backed counterpart profiles from meeting transcripts, on Notion.",
  8. "version": "1.0.0"
  9. },
  10. "plugins": [
  11. {
  12. "name": "ghostwriter",
  13. "description": "Evidence-backed negotiation profiles of the people you meet, kept on Notion and built from meeting transcripts. Prepares meetings and drafts messages. Text only, never sends, every claim backed by a quote.",
  14. "version": "1.0.0",
  15. "author": {
  16. "name": "Elia Rolfi"
  17. },
  18. "license": "MIT",
  19. "category": "productivity",
  20. "source": "./"
  21. }
  22. ]
  23. }
README.md · README · 239 lines
  1. # Ghostwriter
  2. *[Leggi in italiano](README.it.md)*
  3. A Claude Code plugin that keeps an evidence-backed profile of the people you
  4. negotiate with, built from the transcripts of your meetings and stored in your own
  5. Notion workspace.
  6. ## The problem
  7. Between one meeting and the next you forget what people actually care about. Who
  8. needs to justify a decision to whom. What the client said about delivery dates three
  9. months ago. Which concession unblocked things last time. Meeting notes don't help:
  10. they record what was decided, not what each person wants, fears, and will accept.
  11. Ghostwriter reads the transcript of a meeting and keeps, for each counterpart, a
  12. profile that grows over time: drivers, decision style, effective levers, things to
  13. avoid, fears, firm and negotiable positions. **Every line is backed by a verbatim quote
  14. with its timestamp.** When there is no quote, the field stays empty. Before the next
  15. meeting, it tells you what each person risks if they say yes, what to ask, and what
  16. not to say. It drafts the follow-up. It never sends anything.
  17. ## How it works
  18. ```
  19. recording ──► transcript ──► process-meeting ──► Notion: Meeting page + updated profiles
  20. (PLAUD, or (text only, + briefing, recommended move,
  21. any recorder) never audio) follow-up draft, predictions
  22. prepare-meeting · compose-message · consult-profile read the profiles, write nothing
  23. ```
  24. 1. You record a meeting you took part in (a PLAUD device, a phone, anything).
  25. 2. You get a transcript: from PLAUD, from any transcription service, or locally with
  26. `tools/transcribe.py`.
  27. 3. `process-meeting` normalises it into numbered turns, identifies who spoke, extracts
  28. facts, positions and interests, closes the open predictions, and updates each
  29. person's profile on Notion — one person at a time, with a quote for every claim.
  30. 4. `prepare-meeting`, `compose-message` and `consult-profile` only read.
  31. Two Notion databases hold everything: **People** (one page per counterpart) and
  32. **Meetings** (one page per recording, with the transcript in a subpage). The state
  33. of every processing run lives on Notion, so an interrupted run resumes where it
  34. stopped instead of duplicating.
  35. Every briefing closes the open **predictions** and opens new ones, in falsifiable form
  36. (`IF I do X, THEN they will do Y; REFUTED IF Z`). A profile that cannot be proved wrong
  37. cannot be corrected either. A hit rate under 40% means the profile is wrong; above 90%
  38. means the predictions are too obvious.
  39. ## The profile
  40. | Field | What goes in it |
  41. |---|---|
  42. | Key drivers | Status, Security, Control, Autonomy, Recognition, Belonging, Financial result — only once a pattern is seen in 2+ meetings |
  43. | Decision style | Analytical, Intuitive, Consensual, Directive, Avoidant — only after an actual decision was observed |
  44. | Communication style | how they talk, with quotes |
  45. | Goals and interests | what a position lets them do or avoid |
  46. | Fears | risks they perceive, written as topics that need preparation, not weak spots |
  47. | Sensitive topics | business topics to prepare for |
  48. | Effective levers | interest → move → why it works for them → cost to you |
  49. | Things to avoid | what doesn't work or backfires |
  50. | Firm / Negotiable positions | negotiable only if they stated a condition ("if…, then…") |
  51. | Alignment with me | Ally, Favorable, Neutral, Skeptical, Opponent |
  52. | Profile reliability | Low / Medium / High, from the number of meetings with real evidence |
  53. | User notes | yours: the agent reads it, never writes it, and it always wins |
  54. Each claim carries its confidence level (`hypothesis`, `observed`, `pattern`) and the
  55. number of meetings it rests on. Hypotheses never enter the profile fields: they go
  56. into a "to verify" list as open questions. A profile with three fields filled and
  57. eight empty is a healthy profile.
  58. ## What it will never do
  59. - **Analyse audio.** No tone, pitch, hesitation, or emotion inferred from the voice.
  60. Text only. The EU AI Act (art. 5(1)(f)) prohibits inferring emotions from biometric
  61. data such as voice in the workplace; working only on the text is what keeps the
  62. system outside that ban.
  63. - **Send anything.** It writes drafts. An automatic send based on a psychological
  64. profile is exactly what must not be able to happen by mistake.
  65. - **Evaluate the people who report to you.** No performance judgements, no use for
  66. hiring, promotion, duties or discipline. Using it that way would make it a
  67. high-risk system under the AI Act (Annex III) and a workplace-monitoring tool under
  68. most labour laws. People listed as your subordinates get a profile that is mostly
  69. empty by design, and their words are not archived in full.
  70. - **Write anything without a quote.** An empty field beats plausible psychology.
  71. - **Guess who a speaker is.** If it is not certain, it stops and asks: a wrong
  72. attribution poisons two profiles for good.
  73. - **Record special-category data** (health, politics, religion, union membership,
  74. sexual orientation, ethnicity — GDPR art. 9). If it comes up, it is redacted and
  75. never used.
  76. - **Profile people who were not in the room**, or process recordings of conversations
  77. you were not part of.
  78. - **Keep data after a deletion request.** `remove-person` cleans the person out of
  79. every page and tells you the manual steps (trash, version history, local files) that
  80. a real GDPR erasure needs.
  81. The test behind all of it: every line must be one you could show to the person it is
  82. about. The framing is "what does this person risk if they say yes", not "where do I
  83. push".
  84. ## Requirements
  85. - [Claude Code](https://claude.com/claude-code)
  86. - The **Notion MCP connector**, connected to a **personal** Notion workspace (not your
  87. company's: a workspace admin can read everything in it)
  88. - Optional: a **PLAUD** recorder with its MCP connector, to fetch transcripts directly
  89. - Optional: **mlx-whisper** on an Apple Silicon Mac, to transcribe audio locally with
  90. `tools/transcribe.py`
  91. ## Install
  92. As a plugin:
  93. ```
  94. /plugin marketplace add rlfszn/ghostwriter
  95. /plugin install ghostwriter@ghostwriter
  96. ```
  97. Or copy the skills by hand (keep all seven folders together: they reference each
  98. other as siblings):
  99. ```bash
  100. git clone https://github.com/rlfszn/ghostwriter
  101. cp -r ghostwriter/skills/* ~/.claude/skills/
  102. ```
  103. ## First run
  104. Say **"set up Ghostwriter"**. The `setup-notion-memory` skill:
  105. 1. creates your **Ghostwriter home**, `~/.ghostwriter/`, and copies the templates into
  106. it: `configuration.md`, `style.md`, `deletions.md`, plus `transcripts/incoming/` and
  107. `transcripts/archive/`;
  108. 2. asks you to fill in `configuration.md` — at minimum `me` (your name, every alias a
  109. transcript may use for you, a contact email) and `subordinates` (everyone who
  110. reports to you). `me` keeps you from profiling yourself; `subordinates` keeps a
  111. team member from being treated as a counterparty;
  112. 3. checks that the Notion workspace has no other members, then creates the two
  113. databases, their relations, rollups and views, and writes the IDs back into
  114. `configuration.md`.
  115. Your private files live **outside the plugin**, so updates never touch them. For a
  116. separate instance (say, one for work and one personal), run the setup from a project
  117. folder and ask for a per-project instance: it creates `./.ghostwriter/` there, and the
  118. skills use it whenever you work from that folder.
  119. Fill in `style.md` with 3-5 real messages you sent before using `compose-message`,
  120. otherwise the drafts will sound like the agent, not like you.
  121. By default the skills run in **autonomous** mode: they never stop to ask, they leave
  122. anything doubtful marked "to confirm" on Notion. Start a request with
  123. `MODE: interactive` (or put that line in your `CLAUDE.md`) if you want to be asked.
  124. While `confirm_before_writing_profiles` is `true`, interactive runs show you every
  125. profile change before writing it.
  126. ## Usage
  127. | You say | Skill | Result |
  128. |---|---|---|
  129. | "process this morning's meeting" | `process-meeting` | Meeting page + updated profiles + briefing |
  130. | "tomorrow I'm seeing Mark about the budget, how do I approach him?" | `prepare-meeting` | goal, what he risks if he says yes, questions, zero-cost concessions |
  131. | "write me the message for Barlow" | `compose-message` | a draft in your register, to reread |
  132. | "what do we know about Farnsworth?" | `consult-profile` | the profile, and how many meetings it rests on |
  133. | "delete X's profile" | `remove-person` | full clean-up for GDPR, after your confirmation |
  134. | "set up Ghostwriter" / "create the databases on Notion" | `setup-notion-memory` | first-run setup, once |
  135. The skills reply, and write on Notion, in your language. Notion property names stay in
  136. English: they are the contract the queries rely on.
  137. ### Transcribing locally
  138. `tools/transcribe.py` transcribes audio on your Mac without sending it anywhere:
  139. ```bash
  140. uv venv --python 3.11 .venv && uv pip install --python .venv/bin/python mlx-whisper numpy
  141. .venv/bin/python tools/transcribe.py --language en meeting.m4a
  142. ```
  143. The output goes to `transcripts/incoming/` in your Ghostwriter home, without speaker
  144. labels: the agent attributes speakers from the content and says so. With PLAUD and the
  145. speakers renamed in its app, this step is not needed.
  146. ## Using it with other agents
  147. The skills are plain `SKILL.md` markdown files with standard frontmatter. Any agent
  148. that loads skills in that format can use them; there is no code to integrate, only
  149. text to load. Point the agent at `AGENTS.md` for the always-on rules, give it a Notion
  150. connector, and keep the same Ghostwriter home. The external contract is the Notion
  151. schema (`skills/profiling-references/references/notion-schema.md`): anything that
  152. reads or writes the two databases follows it.
  153. The skills were written to run unattended as well (a webhook from your recorder, a
  154. scheduled job). In that mode they never ask, never overwrite a filled field, never
  155. delete, and `remove-person` refuses to run. See
  156. `skills/profiling-references/references/execution-modes.md`.
  157. ## Layout
  158. ```
  159. .claude-plugin/ plugin and marketplace manifests
  160. AGENTS.md always-on rules, for any agent
  161. skills/
  162. profiling-references/ LIBRARY: core rules, method, schema, limits, templates
  163. process-meeting/ the main flow
  164. prepare-meeting/ read-only
  165. compose-message/ read-only + drafts
  166. consult-profile/ read-only
  167. remove-person/ destructive, asks for confirmation
  168. setup-notion-memory/ first-run setup
  169. tools/transcribe.py local transcription with mlx-whisper
  170. tests/ 4 synthetic cases + 8 negative tests
  171. ```
  172. One copy of the logic: the analysis method, the Notion schema and the legal limits
  173. live in the library, and the skills point to them.
  174. ## Tests
  175. `tests/` holds four synthetic meetings with expected results and eight negative tests
  176. where the right behaviour is to stop (insufficient material, noisy transcript,
  177. unknown speakers, sensitive data, deceptive message, single meeting, HR use, voice
  178. analysis). Run them against sandbox databases before your first real recording: say
  179. "run the test suite in the sandbox". Instructions in `tests/README.md`.
  180. ## Legal
  181. This is not legal advice. You are responsible for complying with the laws that apply
  182. to you: in many countries recording a conversation requires telling the other people,
  183. or their consent; workplaces often have policies on recording devices; and data
  184. protection law (the GDPR in the EU) applies to the profiles you keep. Check your
  185. company's policies before recording any meeting — if they forbid it, don't. Tell
  186. people you record, stop when they ask, and if someone asks what you have written about
  187. them, show them the page. `skills/profiling-references/references/ethical-legal-limits.md`
  188. explains the rules the agent applies and the ones that are up to you.
  189. ## License and author
  190. MIT. Created by Elia Rolfi.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.