Skip to content
Skills Library
AI & agents · Learning & explanation

claude-code-automation

/claude-code-automation

Universal reference for Claude Code automation surface — hooks, skills, plugins, sub-agents, agent teams, MCP, scheduled tasks, channels, routines, headless/SDK, checkpointing, permissions. Invoke when designing automations that leverage Claude Code features or when user asks "how do I automate X with Claude Code".

Repository
claude ↗
Licence
MIT
Last repository change
22.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 22.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Universal reference for Claude Code automation surface — hooks, skills, plugins, sub-agents, agent teams, MCP, scheduled tasks, channels, routines, headless/SDK, checkpointing, permissions. Invoke when designing automations that leverage Claude Code features or when user asks "how do I automate X with Claude Code".

Triggers from the description

how do I automate X with Claude Code
Original author description

Universal reference for Claude Code automation surface — hooks, skills, plugins, sub-agents, agent teams, MCP, scheduled tasks, channels, routines, headless/SDK, checkpointing, permissions. Invoke when designing automations that leverage Claude Code features or when user asks "how do I automate X with Claude Code".

How it works

  1. Decision matrix — which surface to use
  2. Hooks — the three lifecycle events that matter most
  3. Skills vs commands vs sub-agents
  4. MCP — internal pattern
  5. Plugins — what to bundle
  6. Scheduled tasks (Desktop)
  7. Headless mode — CI pattern
  8. Permissions — the four-mode model
  9. Memory (memory/ + MEMORY.md)
  10. Checkpointing
Original SKILL.md headings · 127 lines, 703 words
SKILL.md preview
---
name: claude-code-automation
description: Universal reference for Claude Code automation surface — hooks, skills, plugins, sub-agents, agent teams, MCP, scheduled tasks, channels, routines, headless/SDK, checkpointing, permissions. Invoke when designing automations that leverage Claude Code features or when user asks "how do I automate X with Claude Code".
---

# Claude Code Automation

Map of automation surfaces in Claude Code and which one to reach for. Internal-only — relies on what the user already has installed.

## Decision matrix — which surface to use

| Need | Surface | Lives in |
|---|---|---|
| Run shell command on every tool event | **Hook** | `settings.json` `hooks:` |
| Encapsulate domain knowledge + workflow | **Skill** | `.claude/skills/<name>/SKILL.md` |
| Deterministic multi-step flow the user invokes | **Slash command** | `.claude/commands/<name>.md` |
| Isolated worker spawned from main agent | **Sub-agent** | `.claude/agents/<name>.md` |
| Coordinated multi-agent pattern | **Agent team** | `.claude/agent-teams/<name>.yaml` |
| External tool surface (APIs, CLIs) | **MCP server** | `.mcp.json` or `settings.json` `mcpServers:` |
| Package the above for reuse | **Plugin** | `plugins/<name>/.claude-plugin/plugin.json` |
| Runs on a cron | **Scheduled task** (Desktop) or **Routine** (CLI) | Desktop MCP `scheduled-tasks` / `~/.claude/routines/` |
| Non-interactive CI run | **Headless mode** | `claude -p "..." --output-format stream-json` |
| Programmatic embed | **Agent SDK** | `@anthropic-ai/claude-agent-sdk` (TS) or `claude-agent-sdk` (Python) |
| Structured cross-session chat | **Channel** | `channels.json` + `channels-reference` |
| Rich in-CLI UI | **Fullscreen mode** | `claude --fullscreen` or TUI extensions |

## Hooks — the three lifecycle events that matter most

- **PreToolUse** — veto or mutate a tool call. `{"decision":"block","reason":"..."}` blocks.
- **PostToolUse** — react to tool results (auto-format, audit logs).
- **Stop / SubagentStop** — cleanup when work concludes.

```json
{
  "hooks": {
    "PreToolUse": [{
      "matcher": {"tool_name": "Bash"},
      "command": "bash .claude/hooks/protect-sensitive-files.sh"
    }]
  }
}
```

## Skills vs commands vs sub-agents

- **Skills** are knowledge + procedure — auto-invoked by description match. Use for domain expertise.
- **Commands** are deterministic invokable flows — user types `/name`. Use for repeatable operations.
- **Sub-agents** are isolated workers — spawn for parallelism or to protect context window.

## MCP — internal pattern

Prefer MCP tools over shell wrappers when available. Check tool surface first:

```
# From this session
mcp__harness__*           # Harness platform ops
mcp__kubernetes__*        # K8s resources
mcp__9c1c57f5_..._atlassian__*    # Jira + Confluence
mcp__384a610f_..._*       # M365: Outlook, SharePoint, Teams
mcp__MCP_DOCKER__azure_*  # Azure services
mcp__scheduled-tasks__*   # Desktop scheduled tasks
```

## Plugins — what to bundle

A plugin is one coherent outcome: a marketplace item a user can install and get value from. Bundle:
- `.claude-plugin/plugin.json` — metadata.
- `skills/` — auto-invoked knowledge.
- `commands/` — user-invoked flows.
- `agents/` — specialized workers.
- `hooks/` — lifecycle scripts.
- `rules/` — always-loaded instructions.
- `README.md` — what it does + install.

Register in marketplace: edit `<marketplace>/.claude-plugin/marketplace.json`, add a `plugins[]` entry.

## Scheduled tasks (Desktop)

```
mcp__scheduled-tasks__create_scheduled_task
  name: "harness-daily-health"
  cron: "0 9 * * *"
  prompt: "Run /wa-pipeline status for all thelobbi pipelines and post to Slack"
```

## Headless mode — CI pattern

```bash
claude -p "run /wa-report for work-unit-14" \
  --output-format stream-json \
  --allowed-tools "Read,Bash" \
  --max-turns 10
```

Use in Harness CI steps where Claude runs a specific flow without TTY.

## Permissions — the four-mode model

| Mode | When |
|---|---|
| `default` | Interactive — user approves novel calls |
| `acceptEdits` | File edits pre-approved, tools still ask |
| `plan` | Read-only — for analysis |
| `bypassPermissions` | CI / trusted automation only |

Set in `settings.json` `permissions.mode` or per-invocation `--permission-mode`.

## Memory (`memory/` + `MEMORY.md`)

- `MEMORY.md` is the always-loaded index. ≤150 chars per entry, one line.
- Entries written to `memory/<name>.md` with frontmatter (`type: user | feedback | project | reference`).
- Never write memory content into `MEMORY.md` directly.

## Checkpointing

File checkpointing preserves state across sessions. Enable via `settings.json` `"checkpointing": { "enabled": true }`. Use for long-running analyses that must survive restarts.

## Common automation recipes

1. **Auto-format on edit** — PostToolUse hook on Edit/Write + per-language formatter.
2. **Protect secrets** — PreToolUse hook on Bash matching `.env|*.pem|*.pfx|*.key` patterns.
3. **CI gatekeeper** — headless Claude + `--allowed-tools` whitelist + stream-json output parsed by pipeline.
4. **Daily report** — scheduled task runs `/wa-report` + posts to Slack/Teams via MCP.
5. **Work-unit close** — SubagentStop hook emits audit event + triggers report generation.

## Related

- `skills/harness-automation` — CI/CD side.
- `skills/work-unit-protocol` — the reporting contract this automates toward.
- `rules/ultra-mode.md` — constraints every automation respects.

Versions

The database has no other skills named claude-code-automation.

Database snapshot: 13.11.2026 · package version ec64570f4453 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100claude-automation-recommender
    Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code…
  • 100claude-automation-recommender
    Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code…
  • 100claude-automation-recommender
    Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code…
  • 100claude-automation-recommender
    Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code…
  • 100agent-builder
    Build AI agents and automate Claude Code programmatically via the Claude Agent SDK and headless CLI mode. Covers Python SDK, claude -p, SDK MCP servers, hooks, sessions. Triggers on: "build an agent", "agent SDK", "headless mode", "automate Claude", "programmatic agent".
  • 90Claude Code Framework
    Authoritative reference for how Claude Code itself works AND how to build for it — CLAUDE.md memory, custom slash commands, subagents, skills, hooks, MCP servers, settings precedence, permissions, plan mode, CLI/headless/CI, worktrees, and multi-platform surfaces. Use when…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.