Write social
Turn your coding agent into a full product team — composable skills across research, marketing, product, and process. One install, every editor (Claude Code, Cursor, Codex, Windsurf, Gemini CLI, VS Code).
npx -y skills add hungv47/meta-skills --skill write-socialAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 13 stars13 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Platform-native social copy (hook variants + body + CTA + format spec) for tiktok, reels, shorts, x, linkedin. Single-platform per artifact. 5-dim rubric. Not for paid-ad copy (use write-ad), landing-page copy (use write-copy), short-form video briefs with storyboards (use brief-shortform), or long-form articles. Vietnamese polish: polish-vn. AI-tell cleanup: humanmaxxing.
SKILL.md
6.6 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it
Social Copy — Orchestrator
3 agents (copywriter → format-checker → critic) generate platform-native social copy with enforced limits and rubric scoring. Capability: routing.yaml. Agents + dispatch + rubric: references/agent-manifest.md. Methodology: references/playbook.md.
Core question: Does this copy stop the scroll, clear all platform limits, and earn the click — on THIS platform?
Critical Gates — load first
Full detail: references/procedures/critical-gates.md.
- Single-platform per artifact — multi-platform = re-invoke per platform.
- Single-market per artifact — VN auto-routes through
polish-vn. - Brand mode required (
founder/company) — no silent default (Anti-Pattern #5 floor). - Max 1 format-check revision at baseline — two
REVISION_REQUIRED=FORMAT_FAIL.
Before Starting
Apply references/_shared/before-starting-check.md. Then:
| Artifact | Source | Required? |
|---|---|---|
brand/BRAND.md | create-brand | Hard — brand_mode |
research/icp-research.md | research-icp | Rec — audience |
research/product-context.md | research-icp | Rec — voice + proof |
docs/forsvn/artifacts/marketing/brief-shortform/[slug]/brief.md | brief-shortform | Opt — Warm Start |
docs/forsvn/artifacts/marketing/campaign-plan.md | plan-campaign | Opt — cadence |
Mode per references/_shared/mode-resolver.md. budget: standard. --fast → format-check loop = 0; --deep → MAX 2. --fast does NOT skip Cold Start or Critical Gates. Check .forsvn/index/manifest.json for prior marketing/copy/[platform]-*-[slug].md or upstream brief-shortform / plan-campaign. Run Pre-Dispatch per references/procedures/pre-dispatch.md: auto-scan, then 5-Q Cold Start.
Session execution profile (single-vs-multi): inherit per references/_shared/execution-policy.md.
Quality Gate — 5 dimensions
Full rubric + per-platform calibration + Discrimination Test: references/agent-manifest.md § 5-Dim Critic Rubric. Domain rubric: references/rubric.md.
- Hook scroll-stop (Tier 1/2 archetypes per platform-intel §1)
- Char/word limits (hard caps + visible-window per platform-intel §2)
- CTA placement vs algorithm truncation (X/LinkedIn; TikTok/Reels/Shorts default 10)
- Pattern-interruption density (LinkedIn over-density penalty)
- Format compliance (post / thread / carousel / vertical-video caption)
Pass total ≥35/50 AND no dim 0. Discrimination test every cycle.
Artifact Contract
- Path:
docs/forsvn/artifacts/marketing/copy/[platform]-[YYYY-MM-DD]-[slug].md. Lifecycle:pipeline. - Frontmatter (15): verbatim schema (adds legibility
pack_verified+applied_tacticsto the prior 13) →references/format-conventions.md. - Body (in order): Hook variants (
### Variant [A|B|C]) · Body · CTA · Format spec · Legibility block (convention) · Why this works · Critic verdict (6-row table) · Anti-patterns triggered (- Noneif empty).
Side effects + consumed-by + cross-stack contract: references/procedures/artifact-contract.md.
Routing + Dispatch
Single sequential graph (copywriter → format-checker → critic), no branching. Graph + fallback + polish handoff + mode interaction: references/procedures/dispatch-mechanics.md.
Format-check bounce: PASS → critic. REVISION_REQUIRED → copywriter ONCE under ## Format-Checker Feedback — Address Every Violation. Second → FORMAT_FAIL; ships status: blocked; critic NOT dispatched.
Chain Position
Prev: brief-shortform (locks platform/hook/audience/goal) OR plan-campaign (cadence) OR greenfield. Next: see ## Execution. Re-run: new platform, brand-voice shift, hook A/B, underperformance.
Deference: paid ad → write-ad; full video brief → brief-shortform; landing-page → write-copy; long-form (articles, Substack, blog) → write-copy or optimize-seo; VN polish → polish-vn; AI-pattern stripping → humanmaxxing.
Anti-Patterns
references/anti-patterns.md — 14 patterns (10 original + 4 cross-cutting: polish-chain on FAIL/FORMAT_FAIL, multi-platform, VN-market without vn-tone, cross-stack drift). Re-read before ship.
Durable Rules (protected)
<!-- SLOW_UPDATE_START --> <!-- No pinned rules yet. Populate via the slow-update workflow (see references/slow-update-fence.md). Each pinned rule must (a) be procedural not instance-specific, (b) be earned from a regression or critic-flagged failure, (c) cite the artifact / decision record that justified pinning. --> <!-- SLOW_UPDATE_END -->Completion Status
- DONE — copy generated, all format limits passed, critic ≥35, variant_count delivered.
- DONE_WITH_CONCERNS — delivered; critic 25-34 OR any dim <4 OR verdict
fail(FAIL ships with annotations). - BLOCKED — FORMAT_FAIL (two REVISION_REQUIRED); platform unsupported; brief contradicts brand_mode unresolved.
- NEEDS_CONTEXT — no brief, topic, brand voice,
experience/, orbrand/BRAND.md; recommendbrief-shortformorcreate-brand.
Execution
Offer the registry-gated fork (category publish) — Brief-only: polish (humanmaxxing / polish-vn) if flagged; Assisted/Direct need a verified engine. See execution-fork.md; record execution_mode.
Worked Example
LinkedIn founder-voice walkthrough (Cold Start → dispatch → critic 44/50 PASS → polish decision): references/examples/social-walkthrough.md. 10 per-platform strong/weak: references/examples.md.
What ships with it: 55 files
701.1 KB alongside SKILL.md, 8 of them executable
agents/
- copywriter-agent.md12.0 KB
- critic-agent.md10.2 KB
- format-checker-agent.md9.0 KB
references/
- agent-manifest.md4.2 KB
- anti-patterns.md16.7 KB
- examples.md27.7 KB
- examples/social-walkthrough.md13.2 KB
- format-conventions.md12.5 KB
- playbook.md10.6 KB
- procedures/artifact-contract.md4.0 KB
- procedures/critical-gates.md2.3 KB
- procedures/dispatch-mechanics.md13.3 KB
- procedures/pre-dispatch.md12.1 KB
- rubric.md9.8 KB
- _shared/before-starting-check.md9.8 KB
- _shared/execution-fork.md5.6 KB
- _shared/execution-policy.md7.0 KB
- _shared/hook-archetypes.md6.1 KB
- _shared/legibility-convention.md4.9 KB
- _shared/manifest-spec.md29.2 KB
- _shared/marketing-foundations.md7.7 KB
- _shared/mode-resolver.md11.0 KB
- _shared/options-selection.md7.0 KB
- _shared/performance-data.md15.8 KB
- _shared/performance-grounding.md5.8 KB
- _shared/platform-intelligence/CONTRACT.md6.0 KB
- _shared/platform-intelligence/facebook.md9.8 KB
- _shared/platform-intelligence/founder-demo.md6.2 KB
- _shared/platform-intelligence/.generated-support151 B
- _shared/platform-intelligence/linkedin-launch.md15.5 KB
- _shared/platform-intelligence/linkedin.md31.0 KB
- _shared/platform-intelligence/motion-background.md6.7 KB
- _shared/platform-intelligence/newsletter.md12.8 KB
- _shared/platform-intelligence/producthunt.md17.2 KB
- _shared/platform-intelligence/reddit.md11.2 KB
- _shared/platform-intelligence/reels.md29.4 KB
- _shared/platform-intelligence/shorts.md26.0 KB
- _shared/platform-intelligence/showhn.md17.9 KB
- _shared/platform-intelligence/_template.md11.5 KB
- _shared/platform-intelligence/tiktok.md29.4 KB
15 more files not listed here. See all 55 in the repository.