autonomous-loop
SKILL.md
Autonomous Loop
Main orchestration loop that runs continuously until all features are complete.
Quick Start
Start Autonomous Loop
from scripts.autonomous_loop import AutonomousLoop
loop = AutonomousLoop(project_dir)
result = await loop.run()
print(f"Completed: {result.features_completed}")
print(f"Sessions: {result.sessions_used}")
With Configuration
from scripts.loop_config import LoopConfig
config = LoopConfig(
max_sessions=10,
token_budget=500000,
auto_checkpoint=True
)
result = await loop.run(config)
Autonomous Loop Workflow
┌─────────────────────────────────────────────────────────────┐
│ AUTONOMOUS LOOP │
├─────────────────────────────────────────────────────────────┤
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ INITIALIZE │ │
│ │ ├─ Detect session type (INIT vs CONTINUE) │ │
│ │ ├─ Load or create feature list │ │
│ │ └─ Initialize state tracker │ │
│ └────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ FEATURE LOOP │ │
│ │ while (incomplete_features > 0): │ │
│ │ ├─ Select next feature │ │
│ │ ├─ Create checkpoint │ │
│ │ ├─ Implement with TDD │ │
│ │ ├─ Run E2E tests │ │
│ │ ├─ If pass: mark complete │ │
│ │ ├─ If fail: recover or rollback │ │
│ │ └─ Check context limits │ │
│ └────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ CONTEXT CHECK │ │
│ │ if (approaching_limit): │ │
│ │ ├─ Compact context │ │
│ │ ├─ Prepare handoff │ │
│ │ └─ Request continuation │ │
│ └────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ COMPLETION │ │
│ │ ├─ Generate final report │ │
│ │ ├─ Store session memory │ │
│ │ └─ Signal completion │ │
│ └────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────┘
Loop Result
@dataclass
class LoopResult:
success: bool
features_completed: int
features_total: int
sessions_used: int
total_tokens: int
errors_recovered: int
duration_minutes: float
handoff_reason: Optional[str]
Continuation Modes
| Mode | Description | Trigger |
|---|---|---|
| Auto | Loop continues automatically | Context limit |
| Manual | User confirms continuation | Session end |
| Scheduled | Runs at scheduled times | Cron trigger |
| Event | Triggered by events | Git push, CI |
Integration Points
- autonomous-session-manager: Session lifecycle
- coding-agent: Feature implementation
- browser-e2e-tester: Feature verification
- error-recoverer: Handle failures
- checkpoint-manager: Safe rollback
- handoff-coordinator: Session transitions
- progress-tracker: Track and report
References
references/LOOP-LIFECYCLE.md- Loop detailsreferences/CONTINUATION-PROTOCOL.md- Continuation
Scripts
scripts/autonomous_loop.py- Main loopscripts/loop_config.py- Configurationscripts/feature_orchestrator.py- Feature flowscripts/continuation_handler.py- Continuations