← 返回 Skills 市场
lekt9

Foundry

作者 lekt9 · GitHub ↗ · v0.1.0
cross-platform ⚠ suspicious
2870
总下载
0
收藏
8
当前安装
1
版本数
在 OpenClaw 中安装
/install foundry
功能描述
Self-writing meta-extension that forges new capabilities — researches docs, writes extensions, tools, hooks, and skills
使用说明 (SKILL.md)

Foundry

The forge that forges itself. A meta-extension for OpenClaw that researches documentation, learns from failures, and writes new capabilities into itself.

Installation

When the user asks to install Foundry or @getfoundry/foundry-openclaw, run this command:

openclaw plugins install @getfoundry/foundry-openclaw

This will:

  1. Download the plugin from npm
  2. Extract to ~/.openclaw/extensions/foundry/
  3. Enable it automatically
  4. Restart the gateway to load

After installation, the user will have access to all Foundry tools like foundry_research, foundry_write_extension, etc.

Alternative: Manual Config

Add to ~/.openclaw/openclaw.json:

{
  "plugins": {
    "entries": {
      "foundry": { "enabled": true }
    }
  }
}

Configuration

Full configuration options in ~/.openclaw/openclaw.json:

{
  "plugins": {
    "entries": {
      "foundry": {
        "enabled": true,
        "source": "github:lekt9/openclaw-foundry",
        "config": {
          "autoLearn": true,
          "sources": {
            "docs": true,
            "experience": true,
            "arxiv": true,
            "github": true
          },
          "marketplace": {
            "autoPublish": false
          }
        }
      }
    }
  }
}

Config Options

Option Type Default Description
autoLearn boolean true Learn from agent activity automatically
sources.docs boolean true Learn from OpenClaw documentation
sources.experience boolean true Learn from own successes/failures
sources.arxiv boolean true Learn from arXiv papers
sources.github boolean true Learn from GitHub repos
marketplace.autoPublish boolean false Auto-publish high-value patterns

What Foundry Does

Foundry is an AI-powered development agent that can:

  1. Research — Fetch and understand OpenClaw documentation on demand
  2. Write Extensions — Generate new tools and hooks for OpenClaw
  3. Write Skills — Create ClawHub-compatible skill packages
  4. Self-Modify — Add new capabilities to itself
  5. Learn — Record patterns from failures and successes

Tools

Research & Documentation

Tool Description
foundry_research Search docs.openclaw.ai for best practices
foundry_docs Read specific documentation pages

Writing Capabilities

Tool Description
foundry_implement Research + implement a capability end-to-end
foundry_write_extension Write a new OpenClaw extension
foundry_write_skill Write an AgentSkills-compatible skill
foundry_write_browser_skill Write a browser automation skill
foundry_write_hook Write a standalone hook
foundry_add_tool Add a tool to an existing extension
foundry_add_hook Add a hook to an existing extension

Self-Modification

Tool Description
foundry_extend_self Add new capability to Foundry itself
foundry_learnings View learned patterns and insights
foundry_list List all written artifacts

Marketplace

Tool Description
foundry_publish_ability Publish pattern/skill to Foundry Marketplace
foundry_marketplace Search, browse, and install community abilities

Usage Examples

Research before implementing

User: I want to add a webhook to my extension

Agent: Let me research webhook patterns first...
→ foundry_research query="webhook hooks automation"
→ Returns relevant documentation

Now I'll implement it...
→ foundry_add_hook extensionId="my-ext" event="webhook:incoming" ...

Write a new extension

User: Create an extension that monitors GitHub PRs

Agent:
→ foundry_research query="github api webhooks"
→ foundry_write_extension
    id: "github-monitor"
    name: "GitHub Monitor"
    tools: [{ name: "check_prs", ... }]
    hooks: [{ event: "cron:hourly", ... }]

Self-improvement

User: Add a tool that can fetch npm package info

Agent:
→ foundry_extend_self
    action: "add_tool"
    toolName: "foundry_npm_info"
    toolCode: "const res = await fetch(`https://registry.npmjs.org/${p.package}`)..."

How Learning Works

Foundry observes its own tool calls and learns:

  1. Failures → Records error + context
  2. Resolutions → Links fix to failure → Creates pattern
  3. Patterns → Injected as context in future conversations
  4. Crystallization → High-value patterns become permanent capabilities

Security

Foundry validates all generated code before deployment:

  • Blocked: child_process, eval, ~/.ssh, ~/.aws
  • Sandboxed: Extensions tested in isolated process before installation
  • Reviewed: You approve before any code is written to disk

Links

安全使用建议
This skill asks the agent to install a third-party npm plugin that will be extracted into ~/.openclaw and can self-modify and publish new capabilities. The SKILL.md makes safety claims (blocking child_process/eval, sandboxing, user review) but those are unverified. Before installing: 1) Inspect the remote repository and the exact npm package version (github:lekt9/openclaw-foundry) — review the code yourself or have a trusted reviewer do so. 2) Prefer manual installation: run the openclaw plugins install command yourself in a controlled environment, not via an autonomous agent. 3) Back up ~/.openclaw/openclaw.json and disable autoLearn/autoPublish in config (set to false) until you trust the code. 4) If possible, install and test inside an isolated VM/container or ephemeral account. 5) Require and verify any publishing or GitHub/npm tokens; do not grant broad credentials. 6) If you want stricter safety, avoid enabling autonomous invocation for this skill or limit its permissions until you've audited the plugin. If you cannot review the plugin code, treat this skill as high-risk and avoid installing it.
功能分析
Type: OpenClaw Skill Name: foundry Version: 0.1.0 This skill describes a highly capable, self-modifying AI agent that can research, generate, and integrate new code (extensions, tools, hooks, skills) into itself and publish them to a marketplace. While the `SKILL.md` explicitly mentions security controls like blocking `child_process`, `eval`, and access to sensitive directories (`~/.ssh`, `~/.aws`), sandboxing, and human review, the core functionality of arbitrary code generation and self-modification (demonstrated by `foundry_extend_self` with `toolCode` in SKILL.md) presents a significant attack surface and high inherent risk, even if not immediately malicious. The ability to `autoPublish` to an external marketplace (`api.claw.getfoundry.app`) further elevates the risk profile.
能力评估
Purpose & Capability
Name/description (a meta-extension that researches and writes extensions) is coherent with instructions to run 'openclaw plugins install @getfoundry/foundry-openclaw' and to write to ~/.openclaw/extensions/foundry/. Requiring node is reasonable. However, the manifest declares no required config paths or credentials while the SKILL.md instructs writing into ~/.openclaw and mentions publishing to a marketplace — a capability that normally needs tokens/credentials. This mismatch is noteworthy.
Instruction Scope
SKILL.md directs the agent to download and install an npm plugin, edit ~/.openclaw/openclaw.json, and restart the gateway. It also describes network research (docs, GitHub, arXiv) and self-modifying behavior (foundry_extend_self, autoLearn, autoPublish). These are powerful, cross-cutting actions (writing to user config, fetching arbitrary network content, self-modification). The security section claims blocks (child_process, eval, ~/.ssh, ~/.aws) and user review before disk writes, but those are policy claims inside instructions — the skill itself cannot enforce them. The instructions give the agent authority to add code/plugins to the environment, which expands its attack surface.
Install Mechanism
There is no install spec embedded in the skill; instead the SKILL.md tells the agent to install an npm package via 'openclaw plugins install', which will download and extract code from npm into the user's home directory and restart the gateway. Installing a third-party npm package is expected for an extension but is high-risk here because there is no code included for review, no pinned release URL, and the skill's claims about sandboxing and blocking dangerous APIs cannot be validated from this manifest.
Credentials
The registry metadata lists no required environment variables or config paths, yet SKILL.md expects to read/write ~/.openclaw configuration and to potentially publish to a marketplace or access GitHub/arXiv. Publishing or accessing private GitHub/npm/repos typically requires credentials (tokens) that are not declared. The absence of declared credentials while describing networked read/write/publish capabilities is a proportionality mismatch and a red flag.
Persistence & Privilege
always:false (good), but the skill explicitly instructs installing an extension that writes to ~/.openclaw and restarts the gateway and that can self-modify and auto-learn — persistent, system-level changes to the agent environment. These capabilities are powerful; combined with autonomous invocation (platform default) they increase the blast radius. The SKILL.md's promise of user review before disk writes mitigates risk only if actually enforced by the user or platform; it's not enforceable from the instruction file alone.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install foundry
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /foundry 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v0.1.0
Initial public release of Foundry—a self-writing meta-extension for OpenClaw. - Adds `foundry` skill, enabling research, extension generation, self-modification, and automatic learning from successes and failures. - Provides built-in tools for researching documentation, writing extensions/skills/hooks, self-extending, and publishing abilities. - Supports auto-learning from documentation, experience, arXiv, and GitHub sources. - Includes configuration options for enabling/disabling learning sources and marketplace publishing. - Ensures security via code validation, sandboxing, and manual approval for generated code.
元数据
Slug foundry
版本 0.1.0
许可证
累计安装 9
当前安装数 8
历史版本数 1
常见问题

Foundry 是什么?

Self-writing meta-extension that forges new capabilities — researches docs, writes extensions, tools, hooks, and skills. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 2870 次。

如何安装 Foundry?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install foundry」即可一键安装,无需额外配置。

Foundry 是免费的吗?

是的,Foundry 完全免费(开源免费),可自由下载、安装和使用。

Foundry 支持哪些平台?

Foundry 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Foundry?

由 lekt9(@lekt9)开发并维护,当前版本 v0.1.0。

💬 留言讨论