Newsletter VoiceSAFE

An agent skill published by charlie947.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: newsletter-voice
description: >
  Build newsletter writing instructions inside a Codex or Claude project. Runs after voice-builder. Produces newsletter-voice.md, a single file the assistant references when drafting newsletters in the user's voice. Works with or without existing newsletter samples: if the user has past issues, the skill analyses them; if not, the skill offers 6 archetypes tuned to the user's voice. Trigger whenever the user says "build my newsletter voice", "learn my newsletter style", "set up my newsletter system", "train on my newsletters", "newsletter onboarding", or drops newsletter samples into chat asking for an analysis. Requires voice-builder to have run first: the skill needs voice.md and about-me.md in the project to work.
---

# Newsletter Voice

## Codex and Claude runtime

- Use this skill in Codex or Claude with the tools actually available in the current task. `AskUserQuestion` examples describe the questions, not a required API: use an available question tool within its limits, or ask in chat. Reuse answers and source material already supplied.
- Work in the user-selected project. Read its `about-me.md`, `voice.md` and relevant brand files before personalised work. Confirm the intended author if files conflict or contain starter defaults. Ask for missing facts or run `voice-builder`; never inherit the maintainer's identity, accounts or private files.
- Resolve bundled `references/` relative to this skill folder. For an explicitly requested profile refresh, read and update the canonical `about-me.md`, `voice.md` or `newsletter-voice.md` in place, preserving unrelated user facts and rules. Consumers must reread those canonical files. Use a new filename only for new deliverables that would collide with unrelated existing files. Installation alone never starts an interview or writes files. Do not write persistent learnings unless requested.
- Use supplied evidence first. Verify external claims through available search/source tools when needed. If a source or integration is unavailable, name the missing capability and offer supplied text/export input. Never invent facts, first-person experience, metrics or a successful tool run.
- Connect only services needed for the chosen route through the user's existing account. Never print credentials or overwrite connections. Drafting, saving and reviewing do not authorise publishing, sending messages or changing accounts.

## Prerequisites check

The moment this skill is triggered, check the project root for voice.md and about-me.md.

If either file is missing, tell the user:

> Newsletter voice sits on top of your general voice profile. Run voice-builder first (upload the skill or say "build my voice"), then come back here once about-me.md and voice.md are in the project.

Then stop. Do not continue until both files exist.

If both files exist, read them fully, then go straight to Step 1.

## Step 1. Check for samples

Ask the user in chat:

> Do you have 2 to 3 past newsletter issues I can learn from?
>
> Yes: paste them here (one per message or all at once)
> No: type "archetype" and I will build from a template tuned to your voice

Wait for response.

If the user pastes 2 or more newsletters, go to Step 2a.
If the user types "archetype", go to Step 2b.
If the user pastes 1 newsletter, ask for at least one more. If they only have one, offer: "One sample is not enough for pattern detection. Want me to switch to archetype mode and use your one newsletter as a reference point?"

## Step 2a. Sample-based analysis

Read every newsletter fully. Look for patterns across issues, not one-off quirks. Extract:

**Opening formula**
- What the first 3 sentences do (specific result, cultural observation, claim, scene, question)
- Length of the opening section before the first structural break
- Credibility move (how the author establishes authority early)
- Value promise (what the reader is told they will get)

**Section structure**
- Problem or contrast setup
- Named framework or free prose
- Numbered steps, methods, or continuous argument
- Examples and evidence patterns
- Bonus or extension section
- Closing formula and signoff

**Data philosophy**
- Specific numbers per issue (count them)
- Source attribution style (linked, named, uncredited)
- Example-to-abstraction ratio
- Limitation or failure acknowledgements

**Formatting**
- Header usage (frequency, hierarchy)
- List usage (numbered, bulleted, arrows)
- Bold and italic usage
- Prompt, code block, or blockquote formatting
- Visual markers (arrows, checkmarks, emojis if any)

**Length**
- Word count range across samples
- Section word counts

**Voice markers unique to newsletter format**
- Pro tips or callouts (frequency, format)
- Forward-looking closings
- Signoff phrase if consistent across samples
- Meta-transparency (does the author reflect on the process or ask for feedback)

**Absence signals**
- Words, constructions, or structures absent from every sample
- Closing moves the author never uses
- Topics the author never touches

Then go to Step 3.

## Step 2b. Archetype selection

Call the AskUserQuestion tool with a single question:

```json
[
  {
    "question": "Which newsletter archetype fits what you want to write?",
    "header": "Archetype",
    "multiSelect": false,
    "options": [
      {"label": "Data tutorial", "description": "Numbers, frameworks, step-by-step methods with prompts"},
      {"label": "Contrarian essay", "description": "Take a position, defend it, name the opposition"},
      {"label": "Case study teardown", "description": "One subject per issue, unpacked in depth"},
      {"label": "Curated digest", "description": "5 to 7 links with your commentary each week"},
      {"label": "Personal essay", "description": "Reflection on a theme, story-first"},
      {"label": "Interview or profile", "description": "One person per issue, Q and A or narrative"}
    ]
  }
]
```

After the user picks an archetype, load the matching defaults from `

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.

[![ModelsAtlas audit: Newsletter Voice](https://modelsatlas.com/skills/charlie947/newsletter-voice/badge.svg)](https://modelsatlas.com/skills/charlie947/newsletter-voice/)

Not your listing, or something out of date? Claim or update your listing.