Skill Forge

当需要为已有 Skill 自动生成 task_suite.yaml 测试任务集、从 skill_spec.yaml 生成完整 SKILL.md + task_suite、或一键走完「生成 → 评估 → 改进」全链路时使用。 读取 SKILL.md 中的 frontmatter/When to Use/exampl...

This skill appears coherent with its description and has no secret or network access. Before using it: (1) Review any SKILL.md or skill_spec.yaml you pass in to ensure they don't contain secrets (those files' content will be embedded into generated prompts/tasks). (2) Only use --evaluate or --auto-improve if you trust any installed improvement-evaluator and improvement-orchestrator scripts at ~/.claude/skills/ because the CLI will execute those scripts if present. (3) Inspect the generated task_suite.yaml before running it with other tools. If you want extra assurance, run the generator with --mock to avoid LLM calls and avoid the evaluate/auto-improve flags until you vet the local evaluator/orchestrator code.

Type: OpenClaw Skill Name: auto-skill-forge Version: 1.0.1 The auto-skill-forge bundle is a developer utility designed to automate the creation of SKILL.md files and task_suite.yaml test harnesses. The Python scripts (forge.py, task_suite_generator.py, skill_generator.py) use standard libraries to parse YAML/Markdown and generate structured output. While forge.py uses subprocess.run to invoke external evaluation tools located in the user's home directory, it does so using argument lists (avoiding shell injection) and targets specific, expected inter-skill dependencies. No evidence of data exfiltration, malicious persistence, or harmful prompt injection was found.