execution-harnessAgent 执行可靠性体系的导航入口。当用户询问 agent 为什么停了、怎么防 agent 提前退出、execution harness 是什么、40 个 pattern 有哪些、6 轴架构、hook 怎么配时匹配。不执行具体操作——各子 skill 分别处理。
Install via ClawdBot CLI:
clawdbot install lanyasheng/execution-harnessGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Accesses sensitive credential files or environment variables
/etc/passwdPotentially destructive shell commands in tool definitions
rm -rf /Calls external URL not in known-safe list
https://github.com/lanyasheng/execution-harness.gitAI Analysis
The skill appears to be a legitimate productivity/developer tool focused on AI agent reliability patterns. The flagged signals are standard for development tools (git clone, system references) and the rm -rf example is likely illustrative documentation rather than an active destructive command. No evidence of credential harvesting, data exfiltration, or hidden malicious instructions was found.
Generated Apr 24, 2026
SaaS companies using Claude Code for automated code generation and maintenance need to prevent agent early stopping, tool errors, and context loss. The skill ensures persistent execution and error recovery, enabling reliable 24/7 code agent operations.
Enterprises with multiple AI agents handling concurrent code edits require coordination to avoid conflicts and maintain context. This skill provides multi-agent handling patterns, handoff protocols, and quality gates to streamline collaborative development.
Development teams integrating AI agents into CI/CD pipelines can use execution-harness to enforce pre-commit tests, session metrics, and quality gates, ensuring that agent-generated code meets standards before deployment.
Organizations running batch processing or data pipelines with AI agents face rate limits and crashes. The skill's error recovery axis handles rate limits, crash recovery, and MCP disconnections, maintaining task continuity.
AI agents handling complex, multi-step workflows often lose context due to compression or session limits. This skill provides context management, budgeting, and handoff patterns to preserve decision history across sessions.
Offer basic execution-harness hooks for free, with advanced patterns like multi-agent coordination and error recovery as paid upgrades. Scaling based on number of hooks or agent sessions.
Provide tailored execution-harness integration for enterprises needing custom reliability patterns and training. Charge for consulting, custom hook development, and ongoing support.
Sell execution-harness as a premium add-on in AI agent marketplaces (e.g., Claude Code). Users pay per download or subscription for access to reliability patterns.
💬 Integration Tip
Integrate hook scripts into your agent startup by sourcing the relevant axis sub-skills, and configure the bash safety and checkpoint hooks to enforce tool governance. Use the quick start command `bash skills/execution-loop/scripts/ralph-init.sh` to enable persistent execution loops.
Scored Apr 19, 2026
Audited Apr 16, 2026 · audit v1.0
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Systematic code review patterns covering security, performance, maintainability, correctness, and testing — with severity levels, structured feedback guidance, review process, and anti-patterns to avoid. Use when reviewing PRs, establishing review standards, or improving review quality.
Coding style memory that adapts to your preferences, conventions, and patterns for consistent coding.
Provides a 7-step debugging protocol plus language-specific commands to systematically identify, verify, and fix software bugs across multiple environments.
Control and operate Opencode via slash commands. Use this skill to manage sessions, select models, switch agents (plan/build), and coordinate coding through Opencode.
Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions