{
  "markdown": "<p align=\"center\">\n  <img src=\"assets/banner.svg\" alt=\"popcorn-harness banner\" width=\"100%\">\n</p>\n\n<p align=\"center\"><em>Instant on-the-fly harness assembler for Claude Code, Hermes, and OpenClaw.</em></p>\n\n**Popcorn Harness** is a cross-platform plugin that discovers your available skills, agents, and commands on the fly, then pops them together into an optimized execution harness. No manual selection. No guessing. Just describe your task and let it assemble the right capabilities for you.\n\n```\nUser: /popcorn Prep this Next.js app for production\n\n🍿 Popping harness for: Prep this Next.js app for production\nPlatform: Claude Code (project)  |  Tier: 3 — Full Pop\n\nAssembled harness:\n  Phase 1 (parallel): security-review + e2e-testing + seo\n  Phase 2 (sequential): deployment-patterns -> docker-patterns\n\nProceed? [y/n/adjust]\n```\n\n---\n\n## How It Works\n\n1. **Detects platform** — Claude Code, Hermes, or OpenClaw\n2. **Discovers capabilities** — skills, agents, and commands available in your environment\n3. **Applies progressive disclosure** based on task complexity:\n   - **Quick Pop** — 1 capability, executes immediately\n   - **Standard Pop** — 2-3 capabilities, shows plan, confirms\n   - **Full Pop** — 4-5 capabilities or ambiguous task, full confirmation flow\n4. **Assembles execution graph** — parallel phases where possible, sequential where required\n5. **Executes and synthesizes** — structured output with key findings and action items\n\n---\n\n## Platform Support\n\n| Platform | How capabilities are discovered |\n|----------|--------------------------------|\n| Claude Code | `claude agents` CLI + skill dirs (`~/.claude/skills`, `~/.claude/plugins/*/skills`) + command dirs + ECC marketplace (if installed) |\n| Hermes | `available_skills` injected in system prompt |\n| OpenClaw | `available_skills` injected in system prompt |\n\n---\n\n## Installation\n\n### Claude Code — recommended\n\nAdd this repo as a marketplace, then install from it:\n\n```\n/plugin marketplace add seilk/popcorn-harness\n/plugin install popcorn-harness@popcorn-harness\n```\n\n`seilk/popcorn-harness` points to the GitHub repo, which contains `.claude-plugin/marketplace.json`.\nThe second `popcorn-harness` is the marketplace `name` field defined in that file.\n\nAfter installation, run `/reload-plugins` if commands don't appear immediately.\n\n### Claude Code — local development / testing\n\n```bash\nclaude --plugin-dir .\n```\n\nUse `/reload-plugins` after making changes. The local copy takes precedence over any installed version.\n\n### Claude Code — manual install (fallback)\n\n```bash\ngit clone https://github.com/seilk/popcorn-harness ~/.claude/plugins/popcorn-harness\n```\n\n### Hermes\n\n```bash\ncp -r skills/popcorn-harness ~/.hermes/skills/\n```\n\nThe skill will appear in `available_skills` on the next session.\n\n### OpenClaw\n\nAdd the skills directory to `external_dirs` in your Hermes config:\n\n```yaml\n# ~/.hermes/config.yaml\nskills:\n  external_dirs:\n    - /path/to/popcorn-harness/skills\n```\n\n### Team / Custom Marketplace\n\nIf you host your own Claude Code marketplace, add popcorn-harness as a source:\n\n```json\n{\n  \"name\": \"my-marketplace\",\n  \"plugins\": [\n    {\n      \"name\": \"popcorn-harness\",\n      \"source\": { \"source\": \"url\", \"url\": \"https://github.com/seilk/popcorn-harness.git\" }\n    }\n  ]\n}\n```\n\nOr point teammates directly at this repo as a marketplace:\n\n```\n/plugin marketplace add seilk/popcorn-harness\n/plugin install popcorn-harness@popcorn-harness\n```\n\n---\n\n## Usage\n\n### Claude Code\n\n```\n/popcorn <task description>\n```\n\nOr invoke the orchestrator agent directly from any Claude Code session:\n\n```\nUse popcorn-orchestrator to: <task description>\n```\n\n### Hermes\n\n```\npopcorn — <task description>\n```\n\nOr just describe your task and say \"use available skills\":\n\n```\nuse your available skills to review my codebase before deployment\n```\n\n### OpenClaw\n\nSame as Hermes — the skill is automatically available once installed.\n\n---\n\n## Examples\n\n**Single-domain task (Tier 1 — Quick Pop):**\n```\n/popcorn run a security review on this repo\n-> Pops: security-review (1 capability, executes immediately)\n```\n\n**Multi-domain task (Tier 2 — Standard Pop):**\n```\n/popcorn audit my portfolio and suggest rebalancing\n-> Pops: obsidian-finance-vault + tossctl + market-research\n-> Shows plan, waits for confirmation\n```\n\n**Complex task (Tier 3 — Full Pop):**\n```\n/popcorn get this app production-ready\n-> Pops: security-review + e2e-testing + seo + deployment-patterns + docker-patterns\n-> Shows full discovery, proposes phased execution, requires explicit confirmation\n```\n\n---\n\n## Design Principles\n\n- **Progressive disclosure** — reveal complexity only when needed\n- **Evidence-based selection** — capabilities chosen by description match, not name alone\n- **Platform-agnostic** — same skill file works on Claude Code, Hermes, and OpenClaw\n- **Budget-bounded** — hard cap of 5 capabilities per harness run to preserve quality\n- **Fail-transparent** — missing capabilities are reported, not silently skipped\n\n---\n\n## Compatibility\n\n| Component | Minimum version |\n|-----------|----------------|\n| Claude Code | Latest (claude CLI with agent support) |\n| Hermes | Any version with `available_skills` injection |\n| OpenClaw | Any version with `external_dirs` support |\n| ECC (optional) | Any version — enhances discovery if installed |\n\n---\n\n## Troubleshooting\n\n**\"No capabilities found\"**\n- Claude Code: run `claude agents` manually to verify agent discovery works\n- Hermes: check `skills.external_dirs` in `~/.hermes/config.yaml`\n- OpenClaw: verify the skills path is correctly set in config\n\n**\"Platform detected incorrectly\"**\n- Set `OPENCLAW=1` environment variable to force OpenClaw detection\n- Ensure `claude` CLI is in PATH for Claude Code detection\n\n**\"Harness produced poor results\"**\n- Be more specific about what domains to cover (e.g., \"check security, run tests, and generate a deploy checklist\"). More specific tasks produce better capability matching.\n- Manually specify capabilities: \"/popcorn use security-review and e2e-testing to...\"\n\n---\n\n## Contributing\n\nInspired by and built on patterns from:\n- [ECC (Everything Claude Code)](https://github.com/anthropics/everything-claude-code) — `team-builder`, `agent-sort`, `agent-harness-construction`\n\nPRs welcome. Please include examples for any new platform support.\n\n---\n\n## License\n\nMIT\n",
  "bytes": 6330,
  "sha": "4f6fbecd2e343bbe59eee41efc418836cd530ade8e805f426974a8f01374b37d",
  "repo_slug": "seilk/popcorn-harness",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_seilk_popcorn_harness_c07e4e91/readme"
}