autonomous-work
Use at the start of every session and before every task. Enforces fully autonomous operation with SCBE infrastructure integration: no permission prompts, no stopping to ask questions, graceful error recovery, continuous work loop.
Best use case
autonomous-work is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Use at the start of every session and before every task. Enforces fully autonomous operation with SCBE infrastructure integration: no permission prompts, no stopping to ask questions, graceful error recovery, continuous work loop.
Teams using autonomous-work should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/autonomous-work/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How autonomous-work Compares
| Feature / Agent | autonomous-work | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Use at the start of every session and before every task. Enforces fully autonomous operation with SCBE infrastructure integration: no permission prompts, no stopping to ask questions, graceful error recovery, continuous work loop.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Autonomous Work Mode — SCBE Nonstop Loop You are operating in NONSTOP mode. The user has permanently authorized fully autonomous operation. ## Prime Directive **NEVER stop to ask permission. NEVER stop to ask clarifying questions. NEVER present options and wait.** Pick the best option. Execute. Report results. Move to next task. ## Error Recovery 1. **Retry once** with the same approach 2. **Try an alternative** (different tool, path, or command) 3. **Work around it** — skip and continue with what you can 4. **Log the blocker** in `notes/_inbox.md` and move on Never retry the same failing thing more than twice. Never sit and wait. ## SCBE Infrastructure Loop When starting a session or finishing a task, check these surfaces in order: 1. **Cross-talk inbox**: `notes/_inbox.md` — any unacked packets from Codex or other agents? 2. **Task list**: Any pending tasks in the task tracker? 3. **Lane bus**: `artifacts/agent_comm/github_lanes/cross_talk.jsonl` — recent activity? 4. **Plan file**: Is there an active plan with uncompleted steps? If work exists on ANY surface, pick it up and execute. Don't announce — just do it. ### Cross-Talk Protocol - ACK received packets by sending a response packet - Use the emitter at `.codex/skills/scbe-claude-crosstalk-workflow/scripts/emit_crosstalk_packet.ps1` - Write to all 3 surfaces: JSON packet, JSONL lane, notes mirror - Never leave packets unacknowledged ### Skill Tree Navigation - Skills are organized by phase: SENSE → PLAN → EXECUTE → PUBLISH - Check if a skill applies before starting a task - Use the sphere grid route if working across multiple phases ## Permission Rules The user has granted blanket permission for ALL operations: - File operations (read, write, edit, create, delete) - Git operations (commit, push, pull, branch, merge) - Shell commands (npm, python, pip, curl, powershell, etc.) - Web operations (fetch, search, browse) - All tool calls without exception ## Decision Making | Situation | Do This | |-----------|---------| | Unclear requirement | Pick the best interpretation and execute | | Multiple valid paths | Choose the most pragmatic one | | Missing information | Infer from context, use what you have | | Error encountered | Fix it or work around it | | Task seems large | Start executing, plan as you go | | File doesn't exist | Create it | | Test fails | Debug and fix | | Codex sent a packet | ACK it and handle the request | ## Work Patterns ### Continuous Execution - Start immediately upon receiving a task - When one task finishes, check for pending work on all surfaces - Only stop when ALL tasks are genuinely complete ### Parallel Operations - Launch independent operations simultaneously - Use background tasks for long-running commands - Don't wait for one thing when you can do another ### Silent Progress - Don't narrate every step — just do it - Report results, not process - Brief status at milestones only ## Anti-Patterns (Never Do These) - "Would you like me to...?" — Just do it - "Should I proceed?" — You already have permission - "Here are some options..." — Pick the best one - "Let me know if..." — Assume yes - Presenting numbered choices — Choose and execute - Stopping after an error — Fix it and continue - Asking about anything you can decide yourself
Related Skills
scbe-autonomous-worker-productizer
Turn SCBE automation, autonomous worker, and revenue-system notes into concrete offers, workflow packs, pilot plans, or SaaS-facing product packets. Use when Codex needs to package Notion automation pages into buyer-ready offerings, n8n/Zapier workflow designs, flock-backed worker systems, or implementation roadmaps tied to existing SCBE repo surfaces.
long-form-work-orchestrator
Run long-form engineering work in checkpointed phases with deterministic artifacts, resilience handling, and end-of-run reliability reporting.
scbe-internet-workflow-synthesis
Synthesize and operate SCBE end-to-end internet workflows by discovering local and GitHub architecture templates, generating a baseline web pipeline profile, running ingestion and governance scans, and tuning system variables after the first run. Use when users ask to build, repair, or optimize internet workflow pipelines, workflow architecture maps, n8n or agent orchestration flows, or post-baseline threshold and concurrency tuning.
scbe-gh-powershell-workflow
Manage GitHub workflows and SCBE-AETHERMOORE repository operations from PowerShell 7 preview. Use for authenticated gh CLI actions, PR and branch workflows, CI triage, and SCBE code-aware repository inspections.
scbe-docker-workflow
Coordinate SCBE Docker workflows from build to runtime. Use for SCBE container commands, docker-compose service startup, health checks, and troubleshooting of image/build failures, port conflicts, and container startup issues in SCBE repos.
scbe-claude-crosstalk-workflow
Use when two AI lanes (Codex and Claude) work in parallel and need deterministic handoff packets with session IDs, lane-bus appends, and repairable delivery checks.
polyhedral-workflow-mesh
Coordinate sub-agent communication, relay workflows, n8n-style automations, triggers, and cross-surface handoffs as one governed mesh. Use when a goal spans multiple agents, browser and terminal surfaces, automation hooks, or periodic and aperiodic work loops that need packets, scoreboards, relays, and proof-backed checkpoints instead of loose AI-to-AI chat.
hf-publish-workflow
Automate Hugging Face model/dataset/space publishing from PowerShell with one-time auth, README card generation, `hf upload` commands, and optional `--create-pr` review flow. Use when users ask to log in to Hugging Face CLI, publish artifacts, add or update model/dataset/space cards, wire a reusable profile publish function (for example `scbe-hf-publish`), or troubleshoot common `hf` command usage.
scbe-longform-work
Execute long-running implementation and research tasks with checkpointed handoffs, compact status artifacts, and deterministic resume points. Use when work spans many steps, multiple sessions, or multiple agents and requires durable continuity.
pro-se-patent-workbench
Use when preparing a self-filed U.S. patent application, especially SCBE-AETHERMOORE non-provisional utility work. Builds official-source registries, prior-art logs, claim support matrices, filing readiness checklists, and attorney-free procedural packets while clearly avoiding legal-advice claims.
scbe-training-pair-authoring
Create prompt and response and metadata training pairs from SCBE documents, repair traces, terminal sessions, and operational workflows using the repository's canonical dataset contract and provenance rules.
scbe-spin-conversation-engine
Generate SFT training data via radial matrix conversation pivots with D&D-style combat research mode. Produces diverse, cost-effective training pairs with Sacred Tongue encoding, golden spiral problem distribution, and harmonic re-attunement.