From FORSVN
Polishes AI-drafted copy to read human-written: strips AI tells, injects brand voice, compresses text 15%+ with zero idea loss, and optionally runs detector-resistance pass.
How this skill is triggered — by the user, by Claude, or both
Slash command
/forsvn:humanmaxxing [content file or text][content file or text]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Polish-pipeline for **existing** AI-drafted copy: strip AI **tells**, inject brand voice, compress with zero idea **loss** so the text reads **human-written**. 6 sub-agents. Capability metadata: [`routing.yaml`](routing.yaml). Agent table + routes + calibration + detector thresholds: [`references/agent-manifest.md`](references/agent-manifest.md). Methodology: [`references/playbook.md`](referenc...
agents/_template.mdagents/compression-agent.mdagents/critic-agent.mdagents/pattern-scanner-agent.mdagents/soul-injection-agent.mdagents/strip-agent.mdagents/voice-extractor-agent.mdreferences/_shared/before-starting-check.mdreferences/_shared/execution-policy.mdreferences/_shared/mode-resolver.mdreferences/_shared/pre-dispatch-protocol.mdreferences/agent-manifest.mdreferences/ai-patterns.mdreferences/anti-patterns.mdreferences/conciseness-rules.mdreferences/detector-resistance.mdreferences/examples/humanmaxxing-walkthrough.mdreferences/format-conventions.mdreferences/human-writing-stylebook.mdreferences/playbook.mdPolish-pipeline for existing AI-drafted copy: strip AI tells, inject brand voice, compress with zero idea loss so the text reads human-written. 6 sub-agents. Capability metadata: routing.yaml. Agent table + routes + calibration + detector thresholds: references/agent-manifest.md. Methodology: references/playbook.md.
Triggers: "sounds like AI", "too robotic", "humanize this", "ai slop", or terminal call from write-copy/write-ad/write-outreach (Route C). Core question: Would a human editor believe a human wrote this, and cut nothing?
Full text + --fast carve-outs: references/procedures/critical-gates.md.
A single instance = critic auto-FAIL. Full list + enforcement: references/procedures/absolute-prohibitions.md.
Em dashes · "it's not just X, it's Y" · rhetorical-question hooks · colons in prose · "actually" as emphasis · filler context ("in today's…", "rapidly changing") · emojis · unsourced 47/73 · staccato taglines ("Your X, Y'd").
Critic-agent verifies before delivery. Full 10-item checklist + re-dispatch routing: references/procedures/quality-gate.md.
detector_status recorded (not_run/proxy_pass/proxy_fail/pangram_pass/pangram_fail) · Route-C protected tokens verbatimFAIL → re-dispatch named agent (max 2 cycles, then DONE_WITH_CONCERNS).
Apply references/_shared/before-starting-check.md.
| Artifact | Source | Required? |
|---|---|---|
brand/BRAND.md | create-brand | Recommended — voice + lexicon |
research/product-context.md | research-icp | Recommended — voice adjectives + register |
docs/forsvn/artifacts/marketing/content/[slug].md | upstream | Optional — source skill from frontmatter |
docs/forsvn/experience/brand.md | any | Optional — persisted Voice — adjectives |
Canonical Pre-Dispatch (references/_shared/pre-dispatch-protocol.md). Dimensions: target voice, compression (light/moderate/heavy), register, detector mode (none/proxy/pangram), protected tokens (Route C). Warm/Cold Start + write-back: references/procedures/pre-dispatch.md.
Mode (references/_shared/mode-resolver.md) auto-downgrades to A for ≤3 sentences + no prior artifacts. --fast collapses B → A (pattern-scanner + strip + critic), skips detector verification. Never skips Cold Start, Critical Gates, or Prohibitions.
Session execution profile (single-vs-multi): inherit per references/_shared/execution-policy.md.
Routes — A (<200 words, pattern-only) · B (≥200 words, full pipeline) · C (caller-driven). Spawn mechanics: references/procedures/dispatch-mechanics.md.
docs/forsvn/artifacts/marketing/content/[slug].humanmaxxed.md. Route C: embedded in caller's artifact.pipeline — re-run renames prior to [slug].humanmaxxed.v[N].md.skill, version, date, status, compression (%), detector_status, protected_tokens_preserved.polish_chain_applied: humanmaxxing + humanmaxxing_quality_score: N/50 + humanmaxxing_detector_status in their frontmatter.format-conventions.md § "Frontmatter field order" + § "Body section headers".Full template: references/format-conventions.md.
references/anti-patterns.md — 10 pipeline + 4 cross-cutting. Most common: em dash retention, voice-before-strip, surface compression, sterile output.
create-brand or supply samples.Send/publish. On detector fail or DONE_WITH_CONCERNS, re-invoke with critic feedback + tighter compression.
Route B end-to-end (178 → 91 words, 49% compression, 40/50 PASS) + cycle-2 FAIL + --fast + Route C variants: references/examples/humanmaxxing-walkthrough.md.
npx claudepluginhub hungv47/meta-skills --plugin forsvnRewrites AI-generated text to sound natural by removing common AI tells like inflated significance and formulaic phrasing. Matches personal or brand voice from about/me.md or user samples.
Detects and rewrites AI writing patterns (e.g., inflated symbolism, em-dash overuse) to match the user's voice from BRAND.md. Runs on audience-facing assets or ad-hoc text.
Rewrites AI-generated text to sound human by stripping em dashes, stock phrases, promotional hype, and rule-of-three lists. Polish blogs, landing pages, emails, docs after content generation.