claude-world/director-mode-lite / skills/auto-loop
auto-loop
VerifiedTDD-based autonomous development loop with checkpoint recovery and observability changelog
Auto-Loop
Execute a TDD-based autonomous development loop with full observability.
Usage
# Start new task
/auto-loop "Implement user login"
# With acceptance criteria
/auto-loop "Implement authentication
Acceptance Criteria:
- [ ] Login form (email + password)
- [ ] JWT token generation
- [ ] Error handling
"
# Resume interrupted session
/auto-loop --resume
# Force restart (clear old state)
/auto-loop --force "New task"
# Check status
/auto-loop --status
# With iteration limit
/auto-loop "Task" --max-iterations 15
How It Works
┌───────────────────────────────────────────────────────────────┐
│ TDD Iteration │
├───────────┬───────────────────────────────────────────────────┤
│ RED │ Write failing test for next AC │
│ │ → Auto-logged: file_write, test_fail │
├───────────┼───────────────────────────────────────────────────┤
│ GREEN │ Write implementation to make test pass │
│ │ → Auto-logged: file_write/edit, test_pass │
├───────────┼───────────────────────────────────────────────────┤
│ REFACTOR │ Improve code quality (no behavior change) │
│ │ → Use code-reviewer agent for suggestions │
├───────────┼───────────────────────────────────────────────────┤
│ VALIDATE │ Run full test suite + linter │
│ │ → Auto-logged: test_pass/fail │
├───────────┼───────────────────────────────────────────────────┤
│ COMMIT │ Commit changes with descriptive message │
│ │ → Auto-logged: commit │
├───────────┼───────────────────────────────────────────────────┤
│ DECIDE │ Check AC completion → continue or complete │
└───────────┴───────────────────────────────────────────────────┘
Execution
When user runs /auto-loop "<request>":
1. State Detection (Conflict Prevention)
STATE_DIR=".auto-loop"
CHECKPOINT="$STATE_DIR/checkpoint.json"
# Check for existing in-progress session
if [ -f "$CHECKPOINT" ]; then
status=$(jq -r '.status // "unknown"' "$CHECKPOINT" 2>/dev/null || echo "unknown")
iteration=$(jq -r '.current_iteration // 0' "$CHECKPOINT" 2>/dev/null || echo "0")
if [ "$status" == "in_progress" ]; then
echo "⚠️ Found interrupted session at iteration #$iteration"
echo "Options:"
echo " /auto-loop --resume → Continue"
echo " /auto-loop --force \"...\" → Start fresh"
exit 1
fi
fi
Behavior Matrix:
| Existing State | Command | Action |
|---|---|---|
| None | /auto-loop "task" | Start new |
completed | /auto-loop "task" | Archive & start new |
in_progress | /auto-loop "task" | Block - prompt user |
in_progress | /auto-loop --resume | Continue |
in_progress | /auto-loop --force "task" | Archive & start new |
2. Initialize
# Archive old changelog if > 100 lines
CHANGELOG_DIR=".director-mode"
CHANGELOG="$CHANGELOG_DIR/changelog.jsonl"
if [ -f "$CHANGELOG" ] && [ $(wc -l < "$CHANGELOG") -gt 100 ]; then
mv "$CHANGELOG" "$CHANGELOG_DIR/changelog.$(date +%Y%m%d_%H%M%S).jsonl"
fi
# Create state directories
mkdir -p "$STATE_DIR" "$CHANGELOG_DIR"
# Initialize checkpoint
# Build with jq so multiline requests or embedded double quotes stay valid JSON
# (a raw heredoc would break on the quoted acceptance-criteria usage above).
jq -n --arg req "$ARGUMENTS" \
'{request: $req, current_iteration: 0, max_iterations: 20, status: "in_progress", started_at: (now | todate), acceptance_criteria: [], last_test_result: null, files_changed: []}' \
> "$CHECKPOINT"
# acceptance_criteria stays empty here; the parse step below fills it in.
3. Parse Acceptance Criteria
Input:
"Implement authentication
Acceptance Criteria:
- [ ] Login form
- [ ] JWT token
"
Parsed:
{
"acceptance_criteria": [
{ "id": 1, "description": "Login form", "done": false },
{ "id": 2, "description": "JWT token", "done": false }
]
}
4. DECIDE - Completion Check
┌─────────────────────────────────────────────────────────────┐
│ DECIDE - Iteration #3 │
├─────────────────────────────────────────────────────────────┤
│ [x] 1. Login form ← test passing │
│ [x] 2. JWT token ← test passing │
│ [ ] 3. Error handling ← NO TEST YET │
├─────────────────────────────────────────────────────────────┤
│ Decision: 2/3 complete → CONTINUE │
└─────────────────────────────────────────────────────────────┘
Complete when:
- All AC marked
done: true - All tests passing
Stop when:
max_iterationsreached.auto-loop/stopfile exists
Flags
| Flag | Description |
|---|---|
--resume | Continue interrupted session |
--force | Clear old state, start fresh |
--status | Show current session status |
--max-iterations N | Set iteration limit (default: 20) |
Observability
All events are automatically logged via PostToolUse hooks:
| Event | Trigger | Hook |
|---|---|---|
file_write | Write tool | log-file-change.sh |
file_edit | Edit tool | log-file-change.sh |
test_pass/fail | Bash (test) | log-bash-event.sh |
commit | Bash (git commit) | log-bash-event.sh |
Query with /changelog:
/changelog # Recent events
/changelog --summary # Statistics
/changelog --type test # Filter by type
Stop / Resume
# Interrupt (stop after current iteration)
touch .auto-loop/stop
# Check status
/auto-loop --status
# Resume
/auto-loop --resume
# Force restart
/auto-loop --force "New task"
Related
- /changelog - View development events
- code-reviewer - Code quality review
- debugger - Error analysis
- test-runner - Test automation
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 466k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 196k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |