# Using Adept

> Use the `adept` CLI to author an AI skill once and render it into Claude Code, Cursor, Codex, Copilot, OpenCode, and 45+ agents. Apply when installing/syncing skills, editing SKILL.md, or touching .adeptability/.

- **Type:** Skill
- **Install:** `agentstack add skill-itaywol-adeptability-using-adept`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [itaywol](https://agentstack.voostack.com/s/itaywol)
- **Installs:** 0
- **Category:** [Developer Tools](https://agentstack.voostack.com/c/developer-tools)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [itaywol](https://github.com/itaywol)
- **Source:** https://github.com/itaywol/adeptability/tree/main/internal/defaultskills/assets/using-adept
- **Website:** https://pkg.go.dev/github.com/itaywol/adeptability

## Install

```sh
agentstack add skill-itaywol-adeptability-using-adept
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Using the `adept` CLI

`adept` makes one AI skill portable across every coding harness. You **author once** in a
canonical format; `adept` **renders accurately** into each harness's native layout and keeps
both sides in sync.

## Mental model

```
canonical skill  ──render──▶  .claude/skills/…       (per-skill)
(.adeptability/  ──render──▶  .cursor/rules/….mdc    (single-file)
 skills//    ──render──▶  AGENTS.md              (Codex aggregate)
 SKILL.md)       ──render──▶  .github/instructions   (Copilot aggregate)
                 ◀─sync-from─  (adopt edits made directly in a harness file)
```

- **Identity is `(id, content-hash)`** — no version numbers. The hash decides what changed.
- A skill is a directory `/skills//` with one `SKILL.md` (YAML frontmatter +
  markdown body) and optional sidecars (`scripts/`, `references/`, `assets/`).
- An **agent** (subagent) is a single file `.adeptability/agents/.md`; it renders to
  `.claude/agents/`, `.opencode/agents/`, `.cursor/agents/`, `.github/agents/`, and
  `.codex/agents/*.toml` the same way. See [[authoring-adept-agents]].
- The **filesystem is the source of truth.** `config.json` only records enabled harnesses,
  the materialization mode, and library remotes.

## The commands you'll use most

```bash
adept init                       # scaffold .adeptability/ (and seed the default skills)
adept harness add            # enable a harness (claude-code | cursor | codex | copilot | opencode | …)
adept sync                       # render canonical skills → every enabled harness
adept status                     # init state, libraries, harnesses, and drift at a glance
adept diff                       # show exactly what differs between canonical and rendered
adept sync-from                  # adopt edits made directly in a harness file back to canonical
adept skill add  --edit      # scaffold a new skill and open $EDITOR
adept skill install //   # install one skill (pinned to a SHA, safety-scanned)
adept agent add  --edit      # scaffold a new agent (subagent); --template evaluator for reviewers
adept agent check            # safety scan + best-practice lint for an agent
adept loop add  --workflow   # compose a loop: discovery skill + evaluator agent + cron skeleton
```

**`adept --help` is the source of truth for the command surface** — it's always current, so
prefer it over memory. `adept --help` lists every verb; `adept  --help` (e.g.
`adept skill --help`, `adept sync --help`) shows that command's subcommands and flags. Global
flags on every command: `--json`, `--log-level debug|info|warn|error`, `--project `,
`--library `.

## Rules of thumb (so you don't surprise the user)

- **Edit canonical, then `adept sync`.** Don't hand-edit rendered files like
  `.cursor/rules/*.mdc` — they're regenerated. Edited a harness file directly? Run
  `adept sync-from` to pull it back into canonical first.
- **Run `adept status` / `adept diff` before and after a sync** so you can report what changed.
- **Installs are gated.** `adept skill install` runs a safety scan and shows a preview; a
  `critical` finding blocks unless the user passes `--allow-unsafe`. Never pass
  `--allow-unsafe` or `--yes` on the user's behalf without explicit confirmation.
- **Exit codes are meaningful:** `0` clean, `1` error, `2` drift/dirty or merge conflict.
  Scan severities map the same way (`high` → 1, `critical` → 2).
- **Secrets stay in the environment.** Provider API keys are read from the environment at
  call time; adept never writes them to `config.json`.
- Aggregator harnesses (Codex/Copilot) have a **byte budget** — overflow drops the
  lowest-priority skills with a truncation note. Check `adept status`.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [itaywol](https://github.com/itaywol)
- **Source:** [itaywol/adeptability](https://github.com/itaywol/adeptability)
- **License:** MIT
- **Homepage:** https://pkg.go.dev/github.com/itaywol/adeptability

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-itaywol-adeptability-using-adept
- Seller: https://agentstack.voostack.com/s/itaywol
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.