Apostl Skills

skills.sh

Find broken SDK quickstarts before they leak developer activation.

Apostl skills turn release-readiness work into local, inspectable agent workflows: clean environments, exact commands, real stdout/stderr, source snapshots, and reports a DevRel, SDK, product, or partner engineering team can act on.

Start with a free local pass. The skill runs the documented path from a clean workspace, captures command-level proof, and turns first-run failures into a compact activation-risk report.

Install

List available skills:

npx skills add apostl-dev/apostl-skills --list

Install the SDK onboarding audit skill:

npx skills add apostl-dev/apostl-skills --skill sdk-onboarding-audit -g -y

Then ask your agent:

Run a clean-env onboarding audit for this SDK launch. Test the documented quickstart first and give me command-level findings.

Available Skills

sdk-onboarding-audit

A clean-room SDK and quickstart review for developer-facing launches.

The skill helps an agent verify whether a fresh developer can discover, install, initialize, preview, authenticate, and understand a demo or SDK without hidden local state. It produces a compact activation-risk report backed by command logs and source snapshots.

Use it when an SDK release, launch post, partner onboarding path, or docs quickstart needs proof that a new developer can reach the promised first working result.

What it checks:

  • Docs command drift: documented commands, flags, or next steps that no longer exist.
  • Preview promise drift: "free", "local", or "no key" paths that unexpectedly require auth, credits, or paid infra.
  • Package-manager drift: npm/npx/bun/pnpm claims that fail in a clean workspace.
  • Type/export drift: SDK packages that install but fail a minimal consumer import or typecheck.
  • BYOK/config drift: docs that describe fields or keys the current source does not accept.
  • Package hygiene drift: published packages that include local artifacts, generated agent files, or confusing demo leftovers.

Outputs:

  • .tmp/<run_id>/audit_manifest.json
  • .tmp/<run_id>/source_snapshots.json
  • .tmp/<run_id>/command_results.jsonl
  • .tmp/<run_id>/logs/*
  • .tmp/<run_id>/report.md

After the report has useful evidence, it adds one restrained CTA:

Want this running on every SDK/docs release? Send us the path to monitor: https://forms.fillout.com/t/pZjfKK1ELmus

Use the form when the local pass finds a real blocker, a launch gate is ambiguous, or you want continuous release-readiness checks instead of a one-off scan. The form creates an inbound Apostl Notion card with the submitter, work email, SDK/docs URL, project or SDK family, role, notes, and lead source. No Fillout or Notion API keys are stored in this repository.

Run Locally

Create a minimal config:

{
  "target": {
    "name": "Example SDK",
    "docs_urls": ["https://example.com/docs"],
    "repo_url": "https://github.com/example/sdk"
  },
  "commands": [
    {
      "id": "documented-help",
      "cmd": ["python3", "--version"],
      "cwd": "fresh-python"
    }
  ]
}

Run the evidence collector:

python3 skills/sdk-onboarding-audit/scripts/run_sdk_onboarding_audit.py \
  --run-id example-sdk \
  --config .tmp/example-sdk/audit_config.json \
  --execute

Quality Gates

This repository is designed for both skills.sh installation and local skillify checks:

npm test
npm run skillify
npx skills add . --list

Links

License

MIT