agentleFS
imsovikdeimsovikde/souvik-skills/skills/skill-ship/SKILL.md

skill-ship

Skill1 starsChanged 5 months agoMITView on GitHub

Ship any agent skill into the imsovikde/souvik-skills marketplace end to end. Use when the user wants to publish, upload, add, or release a skill to their marketplace; audit or verify whether a skill folder is marketplace-compliant; scaffold a brand-new skill; or adopt a third-party skill into the repository by stripping telemetry, analytics, outbound network calls, and vendor-specific wiring so it becomes a self-contained personal skill. Handles the full repository protocol without needing prior session context: structure, per-skill plugin manifests, every marketplace surface (Claude Code, Cowork, Codex, Antigravity, npm), the Next.js catalog, branch and pull-request flow, owner-only attribution with no AI co-author trailers, verified merges, and the release-please to npm publish chain.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. 0 are common, 5 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create: no skill exists yet. Scaffold one, then ship it.

    line 16
  2. Distinctive:

    Every mode ends by running the preflight script.

    line 21
  3. Distinctive:

    Run this before any commit, in every mode:

    line 25
  4. Distinctive:

    Read every file in the source skill before changing it, including scripts and references.

    line 37
  5. Distinctive:

    Remove vendor-specific install wiring, other marketplaces' manifests, and unrelated CI that does not apply here.

    line 39
  6. Very specific:

    Follow references/skill-ship.md for the exact file list.

    line 46
  7. Very specific:

    Place the skill at skills/<skill-name>/ with SKILL.md, agents/openai.yaml, .claude-plugin/plugin.json, and only the resources it needs.

    line 49
  8. Very specific:

    Update every surface together: .claude-plugin/marketplace.json, codex-plugin.json, README.md, and package.json when a test script is added.

    line 50
  9. Very specific:

    Run npm run validate:skills, npm test, and npm run build.

    line 52
  10. Very specific:

    Commit on a feature branch with Souvik Dey <imsovikde@gmail.com>, a Conventional Commit subject starting feat:, and no Co-Authored-By or …

    line 53

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill Ship

Take a skill from anywhere — a new idea, a folder in another project, or a third-party skill copied off the internet — and land it in imsovikde/souvik-skills correctly, publicly, and verified, with no manual cleanup left for the user.

Read references/skill-ship.md before acting. It is the durable knowledge base for this repository: exact directory shape, every file that must change when a skill is added, the commit and merge protocol, and the release chain. Do not reconstruct these facts from memory or guesswork — session context is routinely lost, and that reference exists so the protocol survives it.

Modes

Pick the mode from what the user has:

  • Create: no skill exists yet. Scaffold one, then ship it.
  • Audit: a skill exists; only report compliance. Change nothing unless asked.
  • Adopt: a third-party skill exists. Make it self-contained and personal, then ship it.
  • Ship: a compliant skill exists. Publish it.

Every mode ends by running the preflight script. Ship and Adopt also run the publish protocol.

Required Preflight

Run this before any commit, in every mode:

node skills/skill-ship/scripts/preflight.cjs <path-to-skill-folder>

It deterministically checks naming, frontmatter, agents/openai.yaml, the per-skill .claude-plugin/plugin.json, forbidden files, hardcoded secrets, and telemetry or outbound-network indicators. Fix every reported failure and re-run until it exits clean. The script reports; you do the fixing — it never edits files.

Adopting A Third-Party Skill

The goal is a skill that runs entirely locally and sends nothing anywhere.

  1. Read every file in the source skill before changing it, including scripts and references.
  2. Remove telemetry, analytics, usage pings, crash/error reporting, remote feature flags, and any hardcoded external endpoint the skill's own workflow does not require. A skill whose actual purpose is calling an API (a search or conversion service) keeps that call — remove tracking, not function. State which calls you kept and why.
  3. Remove vendor-specific install wiring, other marketplaces' manifests, and unrelated CI that does not apply here.
  4. Rewrite SKILL.md frontmatter and body to this repository's standard, and rename the folder to the chosen skill name so folder and name match exactly.
  5. Preserve upstream licence and attribution. If the source carries a licence or NOTICE, keep it inside the skill folder and note the origin in the skill's references/. Stripping telemetry is privacy hardening and is fine; stripping someone's copyright notice is not, and the repository standard already requires preserving legal notices and provenance. If the upstream licence is missing, unclear, or incompatible with MIT redistribution, stop and tell the user before shipping.
  6. Re-run preflight until clean.

Publish Protocol

Follow references/skill-ship.md for the exact file list. In summary:

  1. Confirm GitHub auth is usable, and that the working tree is the real repository (clone it first if running from another project).
  2. Place the skill at skills/<skill-name>/ with SKILL.md, agents/openai.yaml, .claude-plugin/plugin.json, and only the resources it needs.
  3. Update every surface together: .claude-plugin/marketplace.json, codex-plugin.json, README.md, and package.json when a test script is added.
  4. Set the new entry's version to the current package.json version. Release-please owns version bumps; never hand-bump.
  5. Run npm run validate:skills, npm test, and npm run build. All must pass.
  6. Commit on a feature branch with Souvik Dey <imsovikde@gmail.com>, a Conventional Commit subject starting feat:, and no Co-Authored-By or assistant-identity trailers.
  7. Push, open a pull request, and merge it through the GitHub API or web UI so GitHub signs the squash commit and it lands Verified. A direct unsigned push to main is not an acceptable substitute.
  8. Confirm the merged commit reports verified: true and is attributed to imsovikde.
  9. Release-please opens a version/changelog pull request. Merge it too — that fires npm-publish.yml, which publishes to npm and GitHub Packages with provenance.
  10. Confirm the new version is live with npm view @imsovikde/skills version.

Marketplace Coverage

A shipped skill must work across every surface this repository targets. The same SKILL.md standard serves Claude Code, Cowork, Codex, Antigravity, Cursor, and Gemini CLI, so correct structure is what makes it universal — there is no per-marketplace rewrite. references/skill-ship.md lists each surface and the file that carries it.

Verification

Before reporting done:

  • Preflight exits clean.
  • npm run validate:skills, npm test, and npm run build pass.
  • The merged commit is Verified, authored only by the owner, with no AI trailers.
  • The skill appears in marketplace.json, codex-plugin.json, and README.md.
  • The release pull request is merged and the new npm version is live, or the user was told exactly what remains.

Report what ran, what passed, and anything skipped. Never claim a step succeeded without having checked it.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.