Skill 审计报告
agent-launcher-orchestrator 审计报告.
Use when a user wants to build, launch, grade, or schedule a Claude Managed Agent (CMA) in their own Anthropic account — "build me an agent", "launch this as a managed agent", "run this on a schedule", "grade my agent against a rubric", "set up a nightly worker". Reads the per-session goal (./my-agent/goal.json), routes deterministically to one of five phase sub-skills (interview → stage-launch → grade-iterate → run-without-you → wrap-up) via goal_router.py, and compiles the goal+phase into an execution shape (single-pass workflow / bounded grade→iterate loop / recurring cron deployment loop) via loop_compiler.py. Forks context so heavy intake (build sheets, payloads, eval cases) stays out of the parent thread. All launches are emitted as BYOK curl the user runs with their own key; no tool makes API calls. Inspired by anthropics/launch-your-agent (Apache-2.0). Distinct from engineering/agent-harness (generic domain loop) and engineering/write-a-skill (authors Claude Code skills, not CM
OpenAgentSkill 信任评分
OpenAgentSkill 信任评分
Trust Score 帮助 Agent 在安装前判断一个 Skill 是否足以进入候选清单。
GitHub 采用度
通过100
25K 个 GitHub Stars
Star/Fork 活跃度
通过97
25K 个 Star，3.5K 个 Fork; 当前元数据中没有议题活跃度信息
近期维护
通过88
距上次推送 2 个月
许可证清晰度
通过86
MIT
README/SKILL.md 完整度
通过84
元数据包含足够的用法与工作流上下文
依赖与运行时风险
信息64
command execution surface, network or browser surface
安装可用性
通过92
npx skills add alirezarezvani/claude-skills --skill agent-launcher-orchestrator
安装命令安全性
通过92
标准软件包或运行时安装路径
权限范围
警告50
shell or command execution, network or browser access
仓库证据
通过86
https://github.com/alirezarezvani/claude-skills/tree/main/.gemini/skills/agent-launcher-orchestrator
审查状态
信息66
可用 AI 审查数据
Agent 验证结果
信息54
暂未有 Agent 结果数据
检查项
安装与采用审查
安装路径
92
npx skills add alirezarezvani/claude-skills --skill agent-launcher-orchestrator
仓库
88
https://github.com/alirezarezvani/claude-skills/tree/main/.gemini/skills/agent-launcher-orchestrator
许可证
86
MIT
维护
88
距上次推送 2 个月
AI 审查
55
The underlying Python scripts (goal_router.py, loop_compiler.py, goal_state.py) are not included in the submission, so their behavior cannot be fully audited; ensure they do not execute untrusted content from goal.json or make unintended network calls.
README/SKILL.md 完整度
84
Usable description available
依赖风险
64
command execution surface, network or browser surface
安装命令安全性
92
标准软件包或运行时安装路径
权限范围
50
shell or command execution, network or browser access
Star/Fork 活跃度
97
25K 个 Star，3.5K 个 Fork; 当前元数据中没有议题活跃度信息
采用度
88
25K 个 GitHub Stars
警告
- Permission surface may require sandboxing
- The underlying Python scripts (goal_router.py, loop_compiler.py, goal_state.py) are not included in the submission, so their behavior cannot be fully audited; ensure they do not execute untrusted content from goal.json or make unintended network calls.
- The SKILL.md excerpt ends mid-sentence in the forcing-question library, which may indicate truncation in the submitted documentation; verify the full file is present and complete.
- Permission surface needs review: shell or command execution, network or browser access
- Permission surface: shell or command execution, network or browser access
方法
本报告综合公开元数据、AI 审查输出、仓库活跃度、安装就绪度、OpenAgentSkill 事件、质量评分、信任检查和 Agent 安全门槛；它不是完整的源代码安全审计。
对比相近选项
下一步可审计的相关 Skill
mono-color
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
1.9K Stars · 审计报告
Last30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
64K Stars · 审计报告
Academic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
38K Stars · 审计报告