{
  "markdown": "<h1 align=\"center\">harus-skills</h1>\n\n<p align=\"center\">\n  <em>Portable agent skills for durable work state, evidence-backed lessons, and practical tooling.</em>\n</p>\n\n<p align=\"center\">\n  <a href=\"https://github.com/azusachino/harus-skills/actions/workflows/ci.yml\"><img src=\"https://github.com/azusachino/harus-skills/actions/workflows/ci.yml/badge.svg\" alt=\"CI\"></a>\n  <a href=\"LICENSE\"><img src=\"https://img.shields.io/badge/License-MIT-yellow.svg\" alt=\"MIT license\"></a>\n  <img src=\"https://img.shields.io/badge/Agent%20Skills-Standard-blue.svg\" alt=\"Agent Skills Standard\">\n  <img src=\"https://img.shields.io/badge/Works%20with-Claude%20Code%20%7C%20Codex%20%7C%20Antigravity-blueviolet.svg\" alt=\"Works with Claude Code, Codex, and Antigravity\">\n</p>\n\n<p align=\"center\">\n  <sub>Three skills. One graph for memory. One toolbelt for the terminal. Zero runtime code — just markdown an agent reads.</sub>\n</p>\n\n---\n\n## ✨ Skills\n\n| Skill | What it does |\n| --- | --- |\n| 🧠 **`/asobi`** | Durable state across sessions and sub-agents via the [`asobi`](https://github.com/azusachino/asobi) CLI knowledge graph. One graph, four pillars: **session continuity** (`start`/`end`), a **task dispatcher** (`tasks plan\\|list\\|dispatch\\|sync\\|close`) with atomic claims that replaces ephemeral TodoWrite/jsonl, SQLite FTS5/BM25 **keyword recall** (`search` + an ADR log), and a **skill library** (`skills` — install/update skills from git). |\n| 📝 **`/revise`** | Persist lessons, findings, and dead ends so future sessions recall them — positive lessons on the project entity, wrong approaches as active `pitfall` warnings surfaced at the next `/asobi start`. |\n| 🧰 **`/toolbelt`** | Preferred modern CLIs and usage recipes. Mise pins project tools, Make owns tasks, and `jq` handles JSON; the project's declared toolchain takes precedence. |\n\nSkills support the current task and its authorization. They resolve routine uncertainty through evidence, keep task state separate from durable lessons, and verify results in proportion to risk. Current user and project instructions override portable defaults or recalled preferences. Tool availability is checked on the actual machine.\n\nAsobi discovers configuration in ancestor directories, so a nested repository can share its parent's graph. Task dispatch records ownership; the agent still reads relevant lessons and starts any delegated work. Workspaces with a `[skills]` declaration in `asobi.toml` use `asobi skills sync` to reconcile installed skills, including removals.\n\n### 🧰 What's in the toolbelt\n\nModern OSS CLIs, reached for by default over the classic Unix tools:\n\n- **Search & edit** — [`rg`](https://github.com/BurntSushi/ripgrep)/[`fd`](https://github.com/sharkdp/fd) (text/file), [`ast-grep`](https://ast-grep.github.io/) (AST-aware structural search & rewrite), [`sd`](https://github.com/chmln/sd) (find/replace), [`difftastic`](https://github.com/Wilfred/difftastic) (syntax-aware diffs), [`typos`](https://github.com/crate-ci/typos) (source spell-check)\n- **View & inspect** — [`eza`](https://github.com/eza-community/eza)/[`bat`](https://github.com/sharkdp/bat)/[`dust`](https://github.com/bootandy/dust), [`procs`](https://github.com/dalance/procs) (ps), [`doggo`](https://github.com/mr-karan/doggo) (dig), [`hexyl`](https://github.com/sharkdp/hexyl) (hex), [`tailspin`](https://github.com/bensadeh/tailspin) (log highlight)\n- **HTTP & data** — [`xh`](https://github.com/ducaale/xh) (curl), [`dasel`](https://github.com/TomWright/dasel) (YAML/TOML/XML/CSV), [`duckdb`](https://duckdb.org/), [`miller`](https://github.com/johnkerl/miller) (CSV/TSV), [`hyperfine`](https://github.com/sharkdp/hyperfine)/[`oha`](https://github.com/hatoo/oha) (benchmarks)\n- **Runtimes** — [`uv`](https://github.com/astral-sh/uv)/`uvx` (Python), [`bun`](https://github.com/oven-sh/bun)/`bunx` (JS/TS)\n\n## 📦 Installation\n\n### Prerequisites\n\n- A supported agent host (Claude Code, Codex, or Antigravity)\n- [`asobi`](https://github.com/azusachino/asobi) CLI for the `/asobi` skill (`cargo install asobi`)\n\n### Claude Code — Marketplace Plugin\n\n```bash\n/plugin marketplace add azusachino/harus-skills\n/plugin install harus-skills\n```\n\nRestart Claude Code after installing.\n\n### Antigravity (agy) — Plugin\n\n```bash\nagy plugin install https://github.com/azusachino/harus-skills\n# or for local development:\nagy plugin link /path/to/harus-skills\n```\n\n### Codex\n\n```bash\ncodex plugin install https://github.com/azusachino/harus-skills\n```\n\n## 🛠️ Development\n\nThis repository currently uses a Nix devShell and Nix-based CI. The portable tooling default is Mise; migrating this repository's development setup is tracked in the harus-kb remediation plan.\n\n```bash\nnix develop          # Enter dev shell (provides all tools)\nmake fmt             # Format JSON/YAML (not markdown)\nmake check           # Run all checks (format + verify)\nmake validate        # PR gate: check + plugin manifest validation\nmake verify          # Verify repository structure\nmake list-skills     # List all available skills\n```\n\nRun checks explicitly: the current hook installer assumes a directory-backed `.git` and restages files, so it is unsuitable for vendored submodules or partial staging. Manifest validation proves JSON validity and aligned versions; it does not exercise installation in each agent host.\n\n## 🗂️ Skill Structure\n\nEach skill follows the [Agent Skills Standard](http://agentskills.io) format as a flat directory under `skills/`:\n\n```text\nskills/\n  asobi/\n    SKILL.md          # Skill definition with YAML frontmatter\n  revise/\n    SKILL.md\n  toolbelt/\n    SKILL.md\n```\n\n## 🤝 Contributing\n\n1. Create a new directory directly under `skills/`\n2. Add a `SKILL.md` with YAML frontmatter (`name`, `description`, `metadata.version`)\n3. Run `make check` before submitting\n4. Open a pull request\n\n## 🙏 Credits & inspiration\n\nDesign decisions are recorded as [ADRs](docs/adr/). The skills stand on the shoulders of the OSS agent-skill community:\n\n- [**asobi**](https://github.com/azusachino/asobi) — the knowledge-graph CLI that backs `/asobi`\n- [**Agent Skills Standard**](http://agentskills.io) — the `SKILL.md` format every skill targets\n- [**ponytail**](https://github.com/DietrichGebert/ponytail) — reuse existing capabilities and keep solutions small\n- [**karpathy-guidelines**](https://github.com/multica-ai/andrej-karpathy-skills) — surgical changes and goal-driven verification\n\n## 📚 Resources\n\n- [Agent Skills Standard](http://agentskills.io)\n- [Claude Code Documentation](https://code.claude.com/docs)\n",
  "bytes": 6542,
  "sha": "b07464c7d6339f0ccb4cad02d7f0020175af89b039f950cd2a55d74017ce2dc5",
  "repo_slug": "azusachino/harus-skills",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_azusachino_harus_skills_56c728fc/readme"
}