Update to 8gnc — Brand Growth Diagnostic

{
  "name": "voice-profiler",
  "description": "Interview the user about their writing and analyze their real samples to generate a reusable voice profile for the humanize skills. Use when the user asks to build or update a voice profile, identify their writing voice, make future drafts sound more like them, or before humanize when no profile exists.",
  "included_files": [
    {
      "relative_path": "scripts/de_emdash.py",
      "size_in_bytes": 8791
    }
  ],
  "skill_md_contents": "---\nname: voice-profiler\ndescription: Interview the user about their writing and analyze their real samples to generate a reusable voice profile for the humanize skills. Use when the user asks to build or update a voice profile, identify their writing voice, make future drafts sound more like them, or before humanize when no profile exists.\n---\n\n# Voice Profiler — Build Your Writing Voice Profile\n\nAct as a writing voice analyst. Interview the user, analyze their real writing samples, and produce a structured voice profile file that the `humanize` skills can use.\n\n## Why This Matters\n\nGeneric model drafts repeat predictable patterns. `humanize` can edit those patterns, but a voice profile supplies the user's actual rhythm, vocabulary, and preferences. The goal is not generic humanity; it is faithful voice.\n\n## The Interview Process\n\nThis is a conversation, not a form. Adapt based on what the user gives you. Some people will paste 10 writing samples and let you figure it out. Others will want to describe their voice. Both work.\n\n### Phase 1: Collect Writing Samples (Most Important)\n\nAsk the user for 3-5 real writing samples. These are the foundation — everything else is derived from or validated against these.\n\n**What to ask for:**\n- \"Paste 3-5 pieces of writing you're proud of. Emails, LinkedIn posts, proposals, website copy, tweets — anything that sounds like YOU at your best.\"\n- If they have a file with samples: \"Point me to a file with your writing and I'll analyze it.\"\n- If `$ARGUMENTS` contains a file path, read that file for samples.\n\n**What to look for in samples:**\n1. **Sentence rhythm** — What's their natural cadence? Short-long-short? All punchy? Long and flowing?\n2. **Opening patterns** — How do they start pieces? Cold open? Story? Question? Direct statement?\n3. **Closing patterns** — How do they end? Call to action? Reflection? Punchline? Just stop?\n4. **Vocabulary fingerprint** — Words they reach for repeatedly. Phrases that are distinctly theirs.\n5. **Grammar personality** — Fragments? Run-ons? Perfect grammar? Contractions? Formal/informal mix?\n6. **Emotional range** — Do they show humor? Frustration? Vulnerability? Confidence? What's the baseline?\n7. **Persuasion model** — How do they convince? Data? Story? Authority? Diagnosis? Challenge?\n8. **Register shifts** — Do they stay in one gear or shift between casual and formal?\n\n### Phase 2: Guided Questions\n\nAfter analyzing samples, fill gaps with targeted questions. Only ask what you can't already see in the samples.\n\n**Identity & Context:**\n- \"What do you do? How would you describe your role in one sentence?\"\n- \"Who do you write for most often? (clients, colleagues, public audience, specific industry)\"\n\n**Voice Preferences:**\n- \"Are there words or phrases you HATE seeing in writing? Things that make you cringe?\"\n- \"Any words or phrases that are distinctly yours — things people associate with how you communicate?\"\n- \"When you read your own writing back, what makes you think 'yeah, that sounds like me'?\"\n\n**Style Preferences:**\n- \"Do you prefer short and punchy or detailed and thorough?\"\n- \"How do you feel about humor in professional writing?\"\n- \"Do you sign off a specific way? (signature, sign-off phrase, etc.)\"\n\n**Anti-Patterns:**\n- \"What kind of writing makes you physically uncomfortable? Corporate jargon? Fake enthusiasm? Over-qualification?\"\n\n### Phase 3: Synthesis & Validation\n\nBefore generating the profile, play back what you've found:\n\n\"Here's what I see in your voice: [2-3 sentence summary]. Does that sound right, or am I missing something?\"\n\nLet them correct you. The profile should feel like looking in a mirror, not a caricature.\n\n## Output: The Voice Profile File\n\nChoose the destination with the user:\n\n- **Codex with local file access:** prefer `./.8gnc/voice-profile.md` for a project-scoped profile. Offer `~/.config/8gnc/voice-profile.md` only for a user-wide profile and write there only after explicit permission.\n- **ChatGPT or another surface without persistent local file access:** return `voice-profile.md` through the supported downloadable-file workflow and tell the user to save and reattach it when persistence is not confirmed.\n\nNever silently write a personal voice fingerprint into a shared repository.\n\n### Voice Profile Format\n\n```markdown\n---\nname: [Full Name]\nrole: [One-line role description]\ngenerated: [YYYY-MM-DD]\nversion: 1.0\n---\n\n# Voice Profile: [Name]\n\n## Core Identity\n[1-2 sentences describing who this person sounds like. Not what they do — how they SOUND. Use a metaphor or comparison if it fits.]\n\n## Sentence Rhythm\n[Describe their natural cadence pattern. Short-long-short? All punchy? Flowing with sudden stops? Give examples from their samples.]\n\n**Pattern:** [e.g., \"Short-long-short cadence. Punchy declarative → expansion → another punch.\"]\n\n## Opening Patterns\n[How they start pieces. List 2-3 patterns observed in their samples.]\n\n- **[Pattern name]:** [Description + example]\n- **[Pattern name]:** [Description + example]\n\n## Closing Patterns\n[How they end pieces. List 2-3 patterns observed.]\n\n- **[Pattern name]:** [Description + example]\n\n## Vocabulary — Words I Reach For\n[List 15-25 words and phrases that appear repeatedly in their samples or that they identified as distinctly theirs.]\n\n## Vocabulary — Words I Never Use\n[List words and phrases they hate, avoid, or that would sound wrong in their voice. Include common AI words they specifically reject.]\n\n## Persuasion Style\n[How they convince. Name the model and describe it.]\n\n- **Primary:** [e.g., \"Diagnosis model — observe symptom → name it → offer remedy\"]\n- **Secondary:** [e.g., \"Proof before promise — stack evidence before making the ask\"]\n\n## Grammar Personality\n[Their relationship with grammar rules. Fragments? Contractions? Run-ons? Perfect grammar?]\n\n## Emotional Range\n[What emotions show up in their writing and how. Baseline tone + peaks.]\n\n## Register\n[Do they stay in one gear or shift? What triggers a shift?]\n\n## Sign-Off\n[How they sign things. Exact format.]\n\n## Real Writing Samples\n\n### Sample 1: [Context]\n```\n[Paste their actual writing sample]\n```\n\n### Sample 2: [Context]\n```\n[Paste their actual writing sample]\n```\n\n### Sample 3: [Context]\n```\n[Paste their actual writing sample]\n```\n\n## Platform Overrides\n\n### LinkedIn\n[Any LinkedIn-specific voice adjustments — line breaks, hooks, length]\n\n### Email\n[Any email-specific voice adjustments — formality, structure, sign-off]\n\n### Instagram\n[Any IG-specific adjustments — lowercase, density, hashtag style]\n\n### [Other Platform]\n[Add as needed]\n```\n\n## After Generation\n\n1. Write or return the file through the approved path described above.\n2. Tell the user where it was saved and whether the current surface will persist it.\n3. Offer: \"Want to test it? Paste any AI-generated text and I'll run `humanize` using your new profile.\"\n4. Offer: \"Want to add platform-specific overrides? (LinkedIn voice, email voice, Instagram voice)\"\n\n## Post-Processing Note\n\nAfter generating content with `humanize`, run the em dash post-processor when local execution is available. Resolve the installed `voice-profiler` skill directory from this loaded `SKILL.md`, then run:\n\n`python3 [voice-profiler-skill-dir]/scripts/de_emdash.py --max-emdash 2`\n\nIf local execution is unavailable, apply the punctuation rules manually and disclose that the script did not run.\n\nThis replaces excess em dashes with your natural punctuation devices (stdlib-only, no external dependencies). The script's defaults reflect a common human pattern; tune the flags to match what the voice profile observed in the user's samples — `--aside parens|hyphen|comma`, `--elaboration hyphen|comma|colon|semicolon`, `--pivot period|comma|colon`, and `--max-emdash N` (set N to roughly how often the user actually uses em dashes per piece). Run `de_emdash.py --help` for the full flag list.\n\n## Updating an Existing Profile\n\nIf a voice profile already exists at the surface-appropriate path:\n1. Read it first\n2. Ask: \"You already have a voice profile. Want to update it with new samples, or start fresh?\"\n3. If updating: merge new observations with existing profile, keeping what still fits\n4. If starting fresh: run the full interview\n\n## When NOT to Use\n\n- Not for rewriting text — this skill builds the profile; `humanize` and `humanize-ig` consume it.\n- Not for profiling someone else's writing to imitate them — the profile is built from the user's OWN samples; anything else makes every humanized piece sound like the wrong person.\n- Not for inventing a voice from a description alone — push for real samples; descriptions without writing produce a caricature.\n- Not needed every session — run once, update when the user's voice evolves or new samples surface.\n\n## Important Notes\n\n- The profile is only as good as the samples. Push for real writing, not descriptions of writing.\n- 3 samples minimum. 5+ is ideal. More samples = more accurate profile.\n- The user's BEST writing is the target — not their average. Ask for pieces they're proud of.\n- Voice profiles should be updated periodically. Writing voice evolves.\n- Never share the voice profile externally. It's a personal fingerprint.\n- The profile file is plain Markdown — users can edit it manually anytime.\n"
}