# Agent Skill Safety Kit — ProductHunt Launch Pack

A copy-ready launch packet for ProductHunt or similar maker directories. Built for the free **Agent Skill Safety Kit** and its browser-local scorecard / permission manifest validator path.

## Launch decision

**Do not launch from this cron environment unless an authorized ProductHunt account/session is available.** ProductHunt posting is public, account-bound, and often needs logged-in context. This packet exists so Pepe or a future authenticated Jarvis session can launch without rewriting positioning.

## Product

- **Name:** Agent Skill Safety Kit
- **Tagline:** Review AI-agent skills before you install, publish, sell, or delegate them.
- **Website:** https://jarvislandingdeploy.vercel.app/agent-skill-safety-kit.html
- **Free scorecard:** https://jarvislandingdeploy.vercel.app/agent-skill-safety-scorecard.html
- **Permission manifest validator:** https://jarvislandingdeploy.vercel.app/agent-skill-permission-manifest-validator.html
- **Sample mini-report:** https://jarvislandingdeploy.vercel.app/agent-skill-safety-sample-report.html
- **ZIP:** https://jarvislandingdeploy.vercel.app/downloads/agent-skill-safety-kit.zip
- **Category:** Developer Tools / Artificial Intelligence / Security / Productivity
- **Pricing:** Free core kit; optional $29 custom mini-report for high-authority skills.

## Short description

Agent Skill Safety Kit helps builders and teams review reusable AI-agent skills before giving them browser, file, SaaS, memory, customer, or external-action authority. It includes a free scorecard, permission manifest template/schema, browser-local validator, vendor questionnaire, sample mini-report, and marketplace safety checklist.

## Maker comment

Agent skills are starting to behave like software dependencies.

The problem: most reusable skills are distributed as instructions, but the real risk is authority. What can the skill read? What can it write? Which tools can it call? Can it persist memory? What receipt does it leave? How does it fail safely?

I built Agent Skill Safety Kit as a practical review layer before install, sale, listing, or delegation.

What is included:

- a free 100-point browser-local scorecard
- a permission manifest template + JSON schema
- a browser-local manifest validator
- a vendor / marketplace questionnaire
- a sample safety mini-report
- a buyer due-diligence checklist
- a GitHub-ready repo package and downloadable ZIP

It is built for OpenClaw, Claude Code, Hermes, Codex-style agents, browser agents, and internal skill registries.

The goal is simple: before a skill crosses a trust boundary, make its permissions, memory behavior, receipts, exceptions, evals, and rollback gaps visible.

Would love feedback from agent builders, marketplace operators, and teams installing third-party skills.

## Gallery / screenshots

Use the existing hosted preview cards and screenshots in this order:

1. Hero / promise — https://jarvislandingdeploy.vercel.app/media/agent-skill-safety-preview-01.svg
2. Scorecard / review dimensions — https://jarvislandingdeploy.vercel.app/media/agent-skill-safety-preview-02.svg
3. Permission manifest / validator — https://jarvislandingdeploy.vercel.app/media/agent-skill-safety-preview-03.svg
4. Sample mini-report / launch verdict — https://jarvislandingdeploy.vercel.app/media/agent-skill-safety-preview-04.svg

Suggested alt text: “Agent Skill Safety Kit preview showing a scorecard, permission manifest, validator, and sample mini-report for reviewing AI-agent skills before install or launch.”

## Launch checklist

- [ ] Confirm authorized ProductHunt account/session.
- [ ] Use the free kit URL as the primary website.
- [ ] Use the tagline exactly: “Review AI-agent skills before you install, publish, sell, or delegate them.”
- [ ] Upload the four preview SVGs or screenshots in the order above.
- [ ] Paste the maker comment.
- [ ] Pick Developer Tools + Artificial Intelligence + Security where available.
- [ ] Do not lead with the $29 mini-report; mention it only as optional for high-authority skills.
- [ ] After launch, record launch URL in `money-track.md` and add it to future directory/social-proof copy.

## 5 launch-day reply snippets

**What problem does this solve?**
> Agent skills are becoming packages, but most teams only inspect the prompt. The kit reviews the authority layer: tools, data, memory, receipts, exception behavior, evals, and rollback.

**Is it only for OpenClaw?**
> No. OpenClaw is one fit, but the checklist works for Claude Code, Hermes, Codex-style agents, browser agents, and internal skill registries. The core question is runtime-independent: what authority does this skill get?

**Why a permission manifest?**
> Because “this skill helps with CRM enrichment” is not enough. A manifest says exactly what the skill may read, write, call, remember, and prove before it runs.

**Is it free?**
> The kit, scorecard, validator, examples, and templates are free. The $29 mini-report is only for teams/creators who want a custom review for one high-authority skill or workflow.

**Who should use it?**
> Skill creators before publishing, teams before installing third-party skills, marketplaces before listing, and security/platform reviewers before letting skills touch tools or memory.

## Distribution rule

No spam. Share this only in ProductHunt / maker-directory contexts or warm agent-safety discussions where a practical kit is directly relevant.