e8e572e0af
New skills for container agent task orchestration: - /decompose: break tasks into subtasks with dependency graph, write .agent-tasks.json - /orchestrate: check task state, launch container agents in git worktrees, poll for completion Updated with learnings from first real run (agent-runtimes M3, 9 tasks): - Prompts must end with "run pytest and fix failures" - State import conventions explicitly in prompts - Warn about uncommitted files before decomposing (worktrees need committed content) - Copy dependency outputs into downstream worktrees before launching Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
49 lines
2.6 KiB
Markdown
49 lines
2.6 KiB
Markdown
# Custom Claude Skills
|
|
|
|
Reusable [Claude Code](https://docs.anthropic.com/en/docs/claude-code) skills shared across all projects.
|
|
|
|
## Setup
|
|
|
|
```bash
|
|
# Clone and install
|
|
cd ~/dev/claude/custom-claude-skills
|
|
./scripts/install.sh
|
|
```
|
|
|
|
The install script creates symlinks from `~/.claude/skills/` to this repo, making all skills available in every project.
|
|
|
|
## Available Skills
|
|
|
|
| Skill | Invocation | Description |
|
|
|-------|-----------|-------------|
|
|
| reflect | `/reflect M8` | Milestone reflection — reviews conversation history, git log, and project docs to produce structured reflection artifacts across MEMORY.md, FUTURE.md, README.md, and CLAUDE.md |
|
|
| log | `/log` | End-of-session logging — captures decisions, gotchas, open questions to `memory/log/`. Prunes old reflected logs. |
|
|
| reflect-logs | `/reflect-logs` | Processes session logs into topic memory files, deduplicates, flags stale entries. Run periodically. |
|
|
| distill-best-practices | `/distill-best-practices` | Cross-project best practices distillation into `claude-foundations/best-practices/`. Interactive — proposes changes for approval. |
|
|
| context-load | `/context-load` | Reload project context (CLAUDE.md, MEMORY.md, etc.) after `/clear` or mid-session context loss |
|
|
| linter | `/linter` | Audit and manage project formatters/linters. Scan mode checks setup; run mode formats all files. |
|
|
| housekeeping | `/housekeeping` | Cross-project health check: git status, unreflected logs, skill validation, pipeline recommendations |
|
|
| decompose | `/decompose <task>` | Break a task into subtasks with dependency graph. Writes `.agent-tasks.json` for container agent orchestration. |
|
|
| orchestrate | `/orchestrate` | Check task state, launch container agents in git worktrees. Use `/loop 2m /orchestrate` for auto-polling. |
|
|
|
|
## Adding a New Skill
|
|
|
|
1. Create `skills/<skill-name>/SKILL.md` with YAML frontmatter and instructions
|
|
2. Run `./scripts/install.sh` to create the symlink
|
|
3. Update this README's skill table
|
|
|
|
## How Skills Work
|
|
|
|
Skills are Claude Code's extensibility mechanism. Each skill is a markdown file with:
|
|
- **YAML frontmatter** — metadata, tool restrictions, execution mode
|
|
- **Dynamic context injection** — `!`command`` blocks run at load time to pre-fetch data
|
|
- **Structured instructions** — what Claude should do when the skill is invoked
|
|
|
|
Skills in `~/.claude/skills/` are available across all projects (personal scope).
|
|
|
|
## Scripts
|
|
|
|
| Script | Purpose | When to run |
|
|
|--------|---------|-------------|
|
|
| `scripts/install.sh` | Symlinks all skills into `~/.claude/skills/` | After cloning or adding new skills |
|