{
  "markdown": "# skill-builder\n\nPublish and install AI skill prompts from the command line — or author them in the web editor with help from an AI drafting assistant.\n\n```bash\nnpx @dmzagent/skill-builder install @author/skill-id\n```\n\n---\n\n## Getting started: installing skills\n\n### Browse the registry\n\nResults print as a dense table right in your terminal:\n\n```bash\nnpx @dmzagent/skill-builder list\n```\n\nFilter by category, author, type, tag, and sort order:\n\n```bash\nnpx @dmzagent/skill-builder list --category \"Developer Tools\"\nnpx @dmzagent/skill-builder list --type meta          # only meta skills\nnpx @dmzagent/skill-builder list --author skillauthor\nnpx @dmzagent/skill-builder list --sort downloads\nnpx @dmzagent/skill-builder list --json               # machine-readable\nnpx @dmzagent/skill-builder list --web                # open the browser registry instead\n```\n\n### Search and inspect\n\n```bash\nnpx @dmzagent/skill-builder search \"data pipeline\"    # ranked by relevance\nnpx @dmzagent/skill-builder info @skillauthor/dialogue-flow\nnpx @dmzagent/skill-builder suggest dial              # autocomplete: skills, tags, authors\n```\n\nWire up shell completion (skills, tags, and authors complete as you type):\n\n```bash\nnpx @dmzagent/skill-builder completion >> ~/.bashrc\n```\n\n### Install a skill\n\n```bash\nnpx @dmzagent/skill-builder install dialogue-flow\nnpx @dmzagent/skill-builder install @skillauthor/dialogue-flow\n```\n\nDownloads `dialogue-flow.md` and `dialogue-flow.json` to the current directory. Use `--output ./my-skills` to change where files land.\n\n**Meta skills install their dependencies automatically.** Installing a meta skill\nresolves its full dependency tree (deduped, cycle-safe) and installs every\nrequired skill alongside it. Use `--no-deps` to install just the meta skill.\n\n```bash\nnpx @dmzagent/skill-builder install @skillauthor/my-bundle\n# → resolves and installs my-bundle + every skill it depends on\n```\n\n---\n\n## Getting started: publishing skills\n\n### 1. Create an account\n\n```bash\nnpx @dmzagent/skill-builder register myhandle \"My Name\" me@email.com\n```\n\nYou'll be prompted for a password. Or set `SKILL_PASSWORD` to skip the prompt:\n\n```bash\nSKILL_PASSWORD=hunter2 npx @dmzagent/skill-builder register myhandle \"My Name\" me@email.com\n```\n\n### 2. Log in\n\n```bash\nnpx @dmzagent/skill-builder login me@email.com\n```\n\nYour auth token is printed on success. Set it as `SKILL_TOKEN` so subsequent commands can use it:\n\n```bash\nexport SKILL_TOKEN=<token-from-login>\n```\n\n(Or pass `--token $SKILL_TOKEN` to each command.)\n\n### 3. Publish a skill\n\nFrom a markdown file:\n\n```bash\nnpx @dmzagent/skill-builder publish ./my-skill.md\n```\n\nThe CLI reads the filename as the skill id and the first heading as the name. You can also use a JSON manifest:\n\n```bash\nnpx @dmzagent/skill-builder publish ./my-skill.json\n```\n\n### 4. Declare the capabilities a skill needs (optional)\n\nA skill can declare what a model must be able to do to invoke it. The\ndeclaration lives on the spec, so it travels with the skill through the\nregistry, the CLI, and MCP — an agent can tell whether it is able to run a skill\nbefore it spends a turn on it.\n\n```json\n{\n  \"capabilities\": [\n    { \"id\": \"vision\", \"level\": \"required\", \"note\": \"reads chart images\" },\n    { \"id\": \"long-context\", \"level\": \"preferred\" }\n  ]\n}\n```\n\n`required` capabilities gate execution: the registry refuses to run a skill on a\nruntime that cannot provide one, unless the caller explicitly forces a degraded\nrun. `preferred` capabilities only weaken the result. Catalog ids:\n\n```\nvision · audio-input · file-input · structured-output · streaming\nextended-reasoning · long-context · multilingual\ntool-use · parallel-tool-calls · code-execution · web-search · file-system\ncomputer-use · mcp-client\npersistent-memory · citations\n```\n\nCustom ids are accepted for anything the catalog does not model yet. `info` and\n`list` show the contract, and `install` rolls it up across a meta skill's whole\ndependency tree — a meta skill only runs where every skill it installs can run.\n\n```bash\nnpx @dmzagent/skill-builder info @skillauthor/chart-reader\n# Required capabilities (1) — the invoking model must provide these:\n#   - vision (required) — reads chart images\n```\n\n### 5. Fork an existing skill (optional)\n\n```bash\nnpx @dmzagent/skill-builder fork @skillauthor/dialogue-flow --name my-dialogue\n```\n\nCreates a copy under your account that you can republish.\n\n---\n\n## Maintaining the registry: backfill dependencies\n\n`scripts/backfill-dependencies.mjs` finds existing skills that implicitly\norchestrate other skills and proposes explicit `dependencies` + `type: \"meta\"`\nfor them. It reads every skill, detects references to other *real* registry\nskills (by full id, `install` directives, or distinctive names), scores a\nconfidence level, and merges a curated set of known-correct mappings.\n\nIt is **dry-run by default** — it prints a report and writes nothing:\n\n```bash\nnode scripts/backfill-dependencies.mjs                 # report only\nnode scripts/backfill-dependencies.mjs --json          # machine-readable\n```\n\nTo apply, pass `--apply` with a token. Only skills owned by the token's account\nare updated (others are listed and skipped), so run it once per publishing\naccount:\n\n```bash\nSKILL_TOKEN=<token> node scripts/backfill-dependencies.mjs --apply\nnode scripts/backfill-dependencies.mjs --apply --include-medium   # also apply medium-confidence\n```\n\nConfidence tiers: `curated` and `high` (explicit id / install / distinctive-name\nreferences) apply by default; `medium` (fuzzy name match) only with\n`--include-medium`; `meta-no-deps` (orchestration language but no concrete\ndependency) is reported for manual review and never auto-applied.\n\n> Applying requires the registry to run the worker build that understands\n> `type`/`dependencies` (it derives `type` and regenerates markdown on save).\n> The dry-run report works against any deployed version.\n\n---\n\n## Use from coding agents (MCP)\n\nCoding agents — Claude Code, Cursor, Codex, and other MCP clients — can search the\nregistry and **auto-install** skills (resolving meta-skill dependency trees) via the\n[`@dmzagent/skill-builder-mcp`](packages/mcp) MCP server.\n\n```jsonc\n// e.g. .mcp.json (Claude Code) or .cursor/mcp.json (Cursor)\n{\n  \"mcpServers\": {\n    \"skill-builder\": { \"command\": \"npx\", \"args\": [\"-y\", \"@dmzagent/skill-builder-mcp\"] }\n  }\n}\n```\n\nTools: `skill_search`, `skill_info`, `skill_suggest`, `skill_taxonomy`, and\n`skill_install` (writes skill files for the detected tool). See\n[`packages/mcp/README.md`](packages/mcp/README.md) for per-client setup.\n\n---\n\n## Web App\n\nVisit the skill studio at `https://skills.dmzagent.com`. The workspace is three\ncollapsible panes plus a settings drawer:\n\n- **Skill Architect** — an agent that edits the spec directly rather than\n  handing back markdown, with suggested next steps drawn from whatever the spec\n  is still missing.\n- **Spec canvas** — the skill as a document: capability contract, instructions,\n  prompt template with its `{{variables}}` highlighted, examples and tests, and\n  a publish-readiness checklist that links into the field behind each gap.\n- **Preview** — run the skill for real. Template variables become fields, a\n  capability preflight says whether the runtime can honestly execute it, and\n  every run is scored against the expectation it came from. Any run can be\n  handed back to the architect for review.\n- **Architecture** — for meta skills, the dependency tree as a layered graph\n  with unresolved ids and cycles flagged, and the capability contract rolled up\n  across everything the skill installs.\n- **Settings drawer** — every field of the spec, out of the way until you want\n  it: identity, behavior, capabilities, composition, examples and tests, and the\n  generated markdown artifact.\n\n## Install from source\n\n```bash\nnpm install\nnpm run dev --workspace packages/web\n```\n\n## License\n\nMIT\n",
  "bytes": 7887,
  "sha": "6f580b6d7f7090e587463a7918a3524c3490faddd0d931e7508fd09206945dac",
  "repo_slug": "praeceptor-thesis/skill-builder-landing",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_com_dmzagent_skill_builder_mcp_37837b4f/readme"
}