C
Ces107/agent-forge
Free agent skills and Claude Code plugins: launch-check (pre-launch audit of any URL: exposed keys, open…
无 SECURITY.md
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
安装与配置
客户端配置（取自该仓库的安装命令）
{
"mcpServers": {
"ces107-agent-forge": {
"command": "npx",
"args": ["skills"]
}
}
}解析自我们于 2026-10-26 抓取的该仓库 README；我们未执行过。
command/args 来自仓库写明的安装命令；不确定时请以该项目 README 为准。
VS Code（.vscode/mcp.json —— 键名是 servers）
{
"servers": {
"ces107-agent-forge": {
"command": "npx",
"args": ["skills"]
}
}
}VS Code 顶层键是 <code>servers</code>，其余与上面那份相同。
Claude Code（命令行）
claude mcp add --transport stdio ces107-agent-forge -- npx skills
Claude Code 不收 JSON 文件，用这条命令直接加。
配置项的键名以你的客户端为准：客户端与 CLI
- ★ 数取值
- 0 ★ · Ces107/agent-forge · 取自 GitHub 搜索 API · （2026-11-13 20:00:50 核对）
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型：开源项目
- 收录与活跃
- 收录于 2026-10-21 · 最近见到 2026-11-14 · 共见到 28 次 · 被 3 个来源列出
- 评分构成
- 综合分 303 · 3 个来源 · repo
- 安装
npx skills
常见问题（答案都来自我们的记录）
怎么安装？
该仓库给出的安装命令是
npx skills。我们的记录里它是仓库自己提供的；不保证在你的客户端上直接可用。它适合谁 / 用来做什么？
我们把它归在这些标签下: Agent 技能。标签来自条目自身，不是我们强加的分类。
谁维护它、还在更新吗？
仓库
Ces107/agent-forge；我们收录于 2026-10-21；最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。# agent-forge
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
Free agent skills and Claude Code plugins: launch-check (is an AI-built app safe to launch?),
done-gate (no "done" until the tests pass) and four review and verification gates. Plus the
multi-agent pipeline and the payments ledger it built.
## Free skills & plugins
### launch-check
Pre-launch report card for any live URL: security (exposed keys, open Supabase or Firebase tables,
headers), phone design, Google and AI search, speed. Ends with a "fix this first" list.
No install: the hosted free version at https://launch-check-upz2.onrender.com
Claude Code plugin:
/plugin marketplace add Ces107/agent-forge
/plugin install launch-check@agent-forge
Any SKILL.md agent (Codex, Cursor, Copilot, Gemini CLI): npx skills add Ces107/agent-forge --skill launch-check
Details and what it fetches: plugins/launch-check
### done-gate
A Claude Code hook: after a code edit the agent cannot end its turn until a real test command passed
after its last change.
/plugin marketplace add Ces107/agent-forge
/plugin install done-gate@agent-forge
Details, what it runs and how to disable it: plugins/done-gate
### All free skills (this repo, MIT)
Every skill is a SKILL.md folder that runs in Claude Code,
Codex CLI, Cursor, Copilot, Gemini CLI and other agents. The paid ones ship their own tools (scripts
the agent runs), and each has a measured result in its references/how-it-works.md.
bash
/plugin marketplace add Ces107/agent-forge # Claude Code: add this marketplace once
/plugin install agent-forge@agent-forge # Claude Code: all the free skills as one plugin
npx skills add Ces107/agent-forge # any SKILL.md agent
| Skill | What it does | Also on |
| launch-check | New, free, for founders: is your AI-built app safe to launch? One command scores security (exposed keys, open database), phone design, Google & AI search and basics, with a "fix this first" list | free |
| done-gate | New, free: a Claude Code hook that blocks "done" until the tests really ran green after the last edit. Checks the commands that ran, not what the agent says | free |
| proven-code-review | Code review that reads beyond the diff, tries to refute every suspicion, and proves each serious bug with a failing test it ran. Benchmarked in public | Agensi |
| spec-trace-gate | Fails when a spec requirement has no task, a task cites a phantom requirement, or a task names a test that does not exist | Agensi |
| verify-gate | Runs the real tests, coverage, lint, types and security scan; VERIFY: PASS only on all-zero exit codes | Agensi |
| adversarial-review-gate | A review that must break the code: findings count only with a failing test, ending in GATE: PASS/FAIL | Agensi |
Or copy a folder from skills/ into ~/.claude/skills/ (all projects) or .claude/skills/ (one
project). python skills/build.py pre-flights every skill against marketplace scanner rules, keeps
the bundled gate scripts identical to forge/, and writes the zips to dist/skills/.
## The payments ledger
An idempotent double-entry payments ledger that stays exactly-once under concurrent retry storms,
and the multi-agent pipeline that built it. Every correctness claim is checkable from git log in 90
seconds.
## The 90-second proof
done-gate (no "done" until the tests pass) and four review and verification gates. Plus the
multi-agent pipeline and the payments ledger it built.
## Free skills & plugins
### launch-check
Pre-launch report card for any live URL: security (exposed keys, open Supabase or Firebase tables,
headers), phone design, Google and AI search, speed. Ends with a "fix this first" list.
No install: the hosted free version at https://launch-check-upz2.onrender.com
Claude Code plugin:
/plugin marketplace add Ces107/agent-forge
/plugin install launch-check@agent-forge
Any SKILL.md agent (Codex, Cursor, Copilot, Gemini CLI): npx skills add Ces107/agent-forge --skill launch-check
Details and what it fetches: plugins/launch-check
### done-gate
A Claude Code hook: after a code edit the agent cannot end its turn until a real test command passed
after its last change.
/plugin marketplace add Ces107/agent-forge
/plugin install done-gate@agent-forge
Details, what it runs and how to disable it: plugins/done-gate
### All free skills (this repo, MIT)
Every skill is a SKILL.md folder that runs in Claude Code,
Codex CLI, Cursor, Copilot, Gemini CLI and other agents. The paid ones ship their own tools (scripts
the agent runs), and each has a measured result in its references/how-it-works.md.
bash
/plugin marketplace add Ces107/agent-forge # Claude Code: add this marketplace once
/plugin install agent-forge@agent-forge # Claude Code: all the free skills as one plugin
npx skills add Ces107/agent-forge # any SKILL.md agent
| Skill | What it does | Also on |
| launch-check | New, free, for founders: is your AI-built app safe to launch? One command scores security (exposed keys, open database), phone design, Google & AI search and basics, with a "fix this first" list | free |
| done-gate | New, free: a Claude Code hook that blocks "done" until the tests really ran green after the last edit. Checks the commands that ran, not what the agent says | free |
| proven-code-review | Code review that reads beyond the diff, tries to refute every suspicion, and proves each serious bug with a failing test it ran. Benchmarked in public | Agensi |
| spec-trace-gate | Fails when a spec requirement has no task, a task cites a phantom requirement, or a task names a test that does not exist | Agensi |
| verify-gate | Runs the real tests, coverage, lint, types and security scan; VERIFY: PASS only on all-zero exit codes | Agensi |
| adversarial-review-gate | A review that must break the code: findings count only with a failing test, ending in GATE: PASS/FAIL | Agensi |
Or copy a folder from skills/ into ~/.claude/skills/ (all projects) or .claude/skills/ (one
project). python skills/build.py pre-flights every skill against marketplace scanner rules, keeps
the bundled gate scripts identical to forge/, and writes the zips to dist/skills/.
## The payments ledger
An idempotent double-entry payments ledger that stays exactly-once under concurrent retry storms,
and the multi-agent pipeline that built it. Every correctness claim is checkable from git log in 90
seconds.
## The 90-second proof
收录徽章（可选）
想把收录标记放进 README 就复制这行，可选，不影响收录。
[](https://anclum.com/zh/s/ces107-agent-forge/)
- 来源
- github · https://github.com/Ces107/agent-forge
- 按领域浏览
- Agent 技能
- 收录来源
- 3 个独立来源
- 收录于
- 今天
- 安全策略
- 我们没在该仓库根目录找到
SECURITY.md（这不代表不安全）
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。