A specialized skill for decoupling long-running AI coding agents from the orchestrator using tmux for maximum task persistence.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install resilient-coding-agent
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install resilient-coding-agent using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Resilient Coding Agent?
The Resilient Coding Agent skill is designed to manage long-running tasks for AI agents like Codex, Claude Code, and OpenCode. By decoupling the agent process from the primary orchestrator using tmux, it ensures that coding tasks continue to run even if the orchestrator restarts, crashes, or experiences network drops. This architecture provides a robust execution environment for complex engineering tasks that exceed typical session timeouts.
Built with security and reliability in mind, this skill leverages isolated temporary directories and secure prompt handling to prevent shell injection. It is an essential component for developers using Openclaw Skills who require fire-and-forget execution with automated recovery and completion notification systems.
Resilient Coding Agent Use Cases
- Executing complex refactoring tasks that are expected to take more than 5 minutes.
- Running autonomous coding agents in environments where the orchestrator might undergo periodic restarts.
- Managing multiple concurrent coding tasks without losing progress due to process interruptions.
- Monitoring agent progress asynchronously through scrollback capture without being attached to the active process.
How Resilient Coding Agent Works
- The orchestrator creates a secure, unique temporary directory using mktemp to store prompt files and event logs.
- A named tmux session is initialized using a standardized naming convention (e.g., codex-task-name).
- The coding agent is launched within the tmux session, with the task prompt passed securely via file reference rather than shell interpolation.
- For agents like Codex, the skill captures the specific session ID from the event log to enable precise resumption later.
- An optional health monitor loop periodically checks the session status and captures output panes to detect completion or crashes.
- Upon completion or detected failure, the skill either triggers a notification or executes agent-native resume commands to recover the task.
Resilient Coding Agent Setup
Ensure tmux and your preferred coding agent (Codex, Claude Code, OpenCode, or Pi) are installed in your environment. To initialize a resilient session, use the following pattern:
# Create secure temp directory
TMPDIR=$(mktemp -d)
chmod 700 "$TMPDIR"
# Launch the agent in a detached tmux session
tmux new-session -d -s codex-my-task -e "TASK_TMPDIR=$TMPDIR"
tmux send-keys -t codex-my-task 'cd /your/project/dir && codex exec --full-auto "$(cat $TASK_TMPDIR/prompt)" && echo "__TASK_DONE__"' Enter
Resilient Coding Agent Data Schema & Taxonomy
The skill organizes its operational data within a temporary directory ($TMPDIR) to maintain isolation and security:
| File | Description |
|---|---|
prompt |
The raw task instructions written by the orchestrator. |
events.jsonl |
Streaming output and logs from the coding agent. |
codex-session-id |
Extracted thread ID used for resuming specific tasks. |
done |
A marker file indicating successful task completion. |
Resilient Coding Agent Advanced Features
- Secure Prompt Handling: Prevents command injection by referencing prompt files via cat inside double quotes.
- Automated Health Monitoring: Supports active monitoring loops that detect shell prompts or exit indicators to trigger auto-recovery.
- Completion Hooks: Chainable notifications via macOS osascript or system events for immediate feedback.
- Session Decoupling: Survival of tasks through orchestrator restarts by utilizing tmux as a persistent process host.
- Precise Session Recovery: Uses extracted metadata to resume specific agent threads rather than relying on unsafe --last flags.
SKILL.md
Loading
Related Openclaw Skills
Training Manager
A comprehensive workspace management tool for Openclaw Skills designed to scaffold, log, and validate AI agent behaviors and memories.
IPFS Client
A read-only assistant for fetching and exploring IPFS and IPNS content via gateways or local nodes.
Exec Error Doctor
A specialized diagnostic engine designed to triage and resolve command execution failures, shell errors, and API schema mismatches.
swift-patterns
A comprehensive framework for building, reviewing, and refactoring SwiftUI applications using Apple-native best practices and modern API standards.
Japan Finance Agent
A unified MCP agent that aggregates data from six major Japanese financial sources for comprehensive company and macro-economic analysis.
Japan News MCP
A comprehensive AI agent skill for fetching and searching real-time Japanese financial, corporate, and economic news via public RSS feeds.
Featured*
