agentsclimarketplace

Socials

Skill Stepkar2004/init-configurator/src/init_configurator/genome/skills/socials

Agentic project base: Claude Code skills bootstrap and evolve your project, a deterministic Python CLI verifies it. project.yaml manifest, doctor checks, path linting. A project-template alternative for coding agents.

Install
npx -y skills add Stepkar2004/init-configurator --skill socials

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 27 days oldThe repository was created 27 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 1 stars1 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

Publishing and discoverability - fire whenever the user wants to post, announce, launch, or promote anything (a LinkedIn post, an X thread, a Show HN, a launch/reflection draft), wants something findable (GitHub repo invisible in search, topics/description/social preview), or asks for a hook, post draft, or post image/GIF. Runs decide -> optimize -> draft -> post; the human always does the actual posting. Load references/linkedin.md, references/github.md, references/visuals.md on their triggers. Not for long-form video production, and not for in-repo docs (workflow step 4 owns those).

SKILL.md

5.7 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it

socials — decide → optimize → draft → post (the human posts)

Authored 2026-07-11 from a four-source research pass: the longform-factory skill (VideoCreator2, its stage architecture and export format), a live review of real LinkedIn posts, and two web sweeps (LinkedIn algorithm, GitHub discoverability) — raw (2026-07). Platform mechanics rot fast: each platform reference separates evergreen mechanism facts from dated snapshots. Re-verify anything dated older than ~6 months before a high-stakes post.

What the user gates (hard rules)

  • The user picks WHAT to post, from a shortlist — never solo.
  • The user approves final wording, and their edits are canon: a later formatting or export pass must never silently regress words the user touched.
  • The user does the actual posting. NEVER auto-post, never drive a browser or API to publish — the deliverable is a paste-ready package.
  • Posts contain no em dashes and no en dashes (user rule — restructure the sentence instead). Repo docs keep house style; the rule is for post text only.

The workflow

StepWhat happensProduces
1. Decidescore 2-4 candidate angles on the axes below, recommend one, STOP — user pickschosen topic + angle
2. Optimizeload the platform reference; none exists → generic checklist below, offer to research + author oneconstraints the draft must obey
3. Drafthook first (it must fit the platform's fold), preview digest before full text, asset brief via references/visuals.mddraft in docs/posts/
4. Postpaste-ready package (format below), user posts, outcome noted in the draft fileposted + one log line

When the "post" is a page that just needs to be findable (a GitHub repo, a profile), decide and draft collapse — step 2 with the right reference is the whole job.

Decide axes (evidence, not vibes — cite why for each):

  1. Demand proof — has this angle worked for someone else, or does it answer something people actually ask?
  2. Substance density — enough concrete numbers, artifacts, or moments to fill it without padding? If not, the draft will starve.
  3. Audience payoff — who is this for and what do they walk away with?
  4. Timing — is there a hook now (launch, milestone, fresh lesson), or does it keep?

Generic optimize checklist (any platform without a reference)

  • The first visible line carries the whole hook — assume readers never expand.
  • One idea per post. The second idea is the next post.
  • Native format beats a bare external link everywhere; treat any outbound link as a reach cost and decide whether it earns its place.
  • Concrete beats abstract: real numbers, real filenames, real failures.
  • End with a CTA a reader can actually answer, not applause bait.
  • Match the visual to the point (decision table in references/visuals.md).

Voice guardrails (all platforms)

Write for the skim: short lines, no buried subclauses, key numbers on their own line, rounded the way people say them. Second person where natural; concrete metaphor over essay-explainer. Banned: "delve", "landscape", "it's important to note", filler rhetorical questions, engagement bait ("comment YES", "tag someone who"), fake humility openers, generic superlatives, hashtag walls. The test: if a stranger could have written it about any project, cut until only this project's post remains.

Full punctuation and manner mechanics: references/voice.md. A project may add a personal voice.local.md overlay beside it. That overlay is the author's own voice: private, never synced, loaded when present.

The paste-ready package (step 4 format)

Drafts live in docs/posts/YYYY-MM-DD-slug.md (gitignored — drafts are local). Frontmatter: platform, status: draft | approved | posted, date, assets. Body rules, in this order:

  1. Hook options — 2-3 variants, the user picks one (wording gate).
  2. The post — every paste-able field in its own fenced code block (fenced blocks render with click-to-copy); explanation lives between blocks, never inside them.
  3. First comment (only if the link-placement decision put anything there) — own block.
  4. Asset brief or image prompt — own block, per references/visuals.md.
  5. Numbered posting checklist, ending with "report back how it did" — the outcome line goes back into this file and, if a lesson landed, through skill-manager evolve.

Lazy parts (load only on trigger)

TriggerLoad
LinkedIn post: drafting, hooks, timing, link placement, media choicereferences/linkedin.md
GitHub discoverability: repo invisible in search, topics, description, social preview, awesome listsreferences/github.md
Any post visual: explainer image, GIF, carousel, animation, prompt for an image modelreferences/visuals.md
A platform with no reference yet (X, Reddit, HN, ...)generic checklist above; offer to research and author references/<platform>.md per skill-manager authoring standards

What ships with it: 4 files

13.8 KB alongside SKILL.md

references/

Keep looking

Skills are one crate of 327,132. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.