菜单
Use when Promptbeat validate, config inspect, generate, run, eval, report, provider-file loading, environment variables, working_dir, promptfoo, or coding-agent runtime execution fails.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/tophant-ai/aibeat --skill promptbeat-debug-run命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | promptbeat-debug-run |
| description | Use when Promptbeat validate, config inspect, generate, run, eval, report, provider-file loading, environment variables, working_dir, promptfoo, or coding-agent runtime execution fails. |
Diagnose the failing boundary first: project config, provider override, generated promptfoo YAML, target runtime, promptfoo execution, result parsing, or report input.
| Symptom | First check |
|---|---|
missing required --config | The command needs --config <path>. |
missing required --output-dir | promptbeat run and promptbeat eval require --output-dir. |
validate config failure | Run ./bin/promptbeat config inspect --config <promptbeat.yaml> if validation gets far enough. |
working_dir does not exist | Run from repository/release root or use an absolute working_dir. |
| Env template failure | Export the referenced env var; do not paste secrets into YAML. |
| Generated cases exist but eval fails | Check whether the config is Promptbeat project YAML or generated promptfoo YAML. |
| Report fails to parse input | Confirm the input is evaluation_result.json or promptfoo-compatible result JSON. |
| Coding-agent returns no trace | Check adapter trace settings and runtime support. |
promptbeat validate --config <promptbeat.yaml> validates a Promptbeat project.promptbeat generate --config <promptbeat.yaml> --output <cases.json> writes
generated cases only.promptbeat run --config <promptbeat.yaml> --output-dir <dir> runs a
Promptbeat project pipeline.promptbeat eval --config <promptfoo.yaml> --output-dir <dir> runs an existing
promptfoo YAML.promptbeat report --input <evaluation_result.json> renders a report.Use references/common-failures.md for concrete fixes and example commands.
基于 SOC 职业分类
