Token导航 LogoToken导航

Agent Skill · 研究检索

schemelike-metacircular-eval(方案式元循环评估)

schemelike-metacircular-eval 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

schemelike-metacircular-eval 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为执行命令;许可证为 MIT,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 46,227。

总安装 792周安装 33GitHub Stars 93下载 264核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:schemelike-metacircular-eval(方案式元循环评估)
来源仓库:https://github.com/letta-ai/skills
仓库路径:skills/schemelike-metacircular-eval
安装命令:
npx skills add https://github.com/letta-ai/skills --skill schemelike-metacircular-eval
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/letta-ai/skills --skill schemelike-metacircular-eval

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

schemelike-metacircular-eval 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果时使用。它可结合来源仓库、安装命令和原始 README 继续核验具体用法。

安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。当前顶部介绍为空,需参考原始 SKILL.md 获取详细功能说明。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/schemelike-metacircular-eval

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/letta-ai/skills --skill schemelike-metacircula…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要执行命令,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

schemelike-metacircular-eval 怎么安装?

在终端执行 npx skills add https://github.com/letta-ai/skills --skill schemelike-metacircular-eval。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

schemelike-metacircular-eval 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

schemelike-metacircular-eval 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/letta-ai/skills --skill schemelike-metacircular-eval;npx skills add letta-ai/skills --skill "schemelike-metacircular-eval" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

schemelike-metacircular-eval 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

schemelike-metacircular-eval 用的人多吗?

公开数据是累计安装 792 次,近一周安装 33 次,下载 264 次,GitHub 93 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

schemelike-metacircular-eval 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
schemelike-metacircular-eval

方案式元循环评估 · 本页

79293执行命令MIT通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Schemelike Metacircular Eval

Overview

This skill provides guidance for building metacircular evaluators—interpreters written in the same language they interpret, capable of interpreting themselves. These tasks require careful handling of evaluation levels, environment structures, and the distinction between host-level and interpreted-level data.

Initial Exploration Phase

Before writing any code:

  1. Examine the host interpreter implementation - Read the existing interpreter (e.g., interp.py) to understand what primitives are provided, how values are represented, and what the evaluation model looks like.
  2. Study the test files - Understand what constructs need to be supported by examining test cases. Identify which tests involve single-level vs. multi-level interpretation.
  3. Establish reliable test execution - Verify the testing mechanism works correctly before deep debugging. Shell command quirks (like echo -e vs printf behavior) can cause false positives in debugging.
  4. Create a mental model of interpretation levels:

- Level 0: The host language (e.g., Python running interp.py) - Level 1: The first interpreter (eval.scm running in Python) - Level 2: The metacircular layer (eval.scm interpreting eval.scm)

Core Implementation Strategy

Environment Management

Implement environment operations with clear semantics:

;; Environment structure: list of frames, each frame is list of (name . value) pairs
(define (make-env) '(()))

(define (extend-env params args env)
  ;; Create new frame with parameter bindings, prepend to env
  (cons (make-frame params args) env))

(define (lookup var env)
  ;; Search frames from innermost to outermost
  ...)

(define (env-define! var val env)
  ;; Add binding to current (first) frame
  ...)

Critical: Test environment operations in isolation before using them in the evaluator. Create unit tests for:

  • Empty environment lookup (should error)
  • Single binding lookup
  • Shadowed variable lookup
  • Definition in nested environments

Closure Representation

Closures must capture their defining environment:

(define (make-closure params body env)
  (list 'closure params body env))

(define (closure? obj)
  (and (pair? obj) (eq? (car obj) 'closure)))

(define (closure-params c) (cadr c))
(define (closure-body c) (caddr c))
(define (closure-env c) (cadddr c))

Key insight for metacircular interpretation: At level 2, closures are data structures that the level-1 interpreter must correctly recognize and apply. Ensure predicates like closure? work on data that has passed through multiple interpretation layers.

The Eval-Apply Loop

(define (eval exp env)
  (cond
    ((self-evaluating? exp) exp)
    ((variable? exp) (lookup exp env))
    ((quoted? exp) (cadr exp))
    ((definition? exp) (eval-definition exp env))
    ((if? exp) (eval-if exp env))
    ((lambda? exp) (make-closure (lambda-params exp) (lambda-body exp) env))
    ((let? exp) (eval-let exp env))
    ((begin? exp) (eval-sequence (begin-actions exp) env))
    ((application? exp) (apply-proc (eval (car exp) env)
                                     (eval-args (cdr exp) env)
                                     env))
    (else (error "Unknown expression type"))))

(define (apply-proc proc args env)
  (cond
    ((primitive? proc) (apply-primitive proc args))
    ((closure? proc)
     (eval-sequence (closure-body proc)
                    (extend-env (closure-params proc)
                                args
                                (closure-env proc))))
    (else (error "Unknown procedure type"))))

Handling let Expressions

Two approaches:

Direct implementation:

(define (eval-let exp env)
  (let ((vars (let-vars exp))
        (vals (map (lambda (e) (eval e env)) (let-vals exp)))
        (body (let-body exp)))
    (eval-sequence body (extend-env vars vals env))))

Transform to lambda (fallback approach):

(define (let->lambda exp)
  (cons (list 'lambda (let-vars exp) (let-body exp))
        (let-vals exp)))

Note: Transforming let to lambda can be a workaround but may mask underlying issues. Prefer direct implementation and investigate if problems persist.

Debugging Multi-Level Interpretation

Systematic Bisection Approach

When something works at level 1 but fails at level 2:

  1. Identify the failing component - Is it environment lookup, closure application, or expression evaluation?
  2. Create minimal reproduction - Write the smallest program that demonstrates the failure: ;; Test at level 1 (eval '(let ((x 1)) x) (make-env));; Test at level 2 - have eval.scm interpret a simple let
  3. Add debug output - Insert print statements at key points:

- Before/after environment extension - When creating closures - When applying procedures - When looking up variables

  1. Compare execution traces - Run the same expression at level 1 and level 2, comparing the debug output to find where behavior diverges.

Common Multi-Level Issues

Predicate confusion: At level 2, pair? on the host might return #t for a closure structure, but the interpreted pair? might behave differently if closures are represented as lists.

Environment corruption: When environments pass through multiple interpretation layers, ensure the structure is preserved. A common bug is the environment becoming "flattened" or losing nested structure.

Primitive vs. closure distinction: Ensure primitive? and closure? predicates are mutually exclusive and correctly identify procedures at all interpretation levels.

Testing Strategy

Unit Test Core Abstractions First

Before testing full programs, verify:

;; Test environment operations
(define test-env (make-env))
(env-define! 'x 1 test-env)
(assert (= (lookup 'x test-env) 1))

;; Test shadowing
(define nested-env (extend-env '(x) '(2) test-env))
(assert (= (lookup 'x nested-env) 2))

;; Test closure creation and access
(define test-closure (make-closure '(a) '((+ a 1)) test-env))
(assert (closure? test-closure))
(assert (equal? (closure-params test-closure) '(a)))

Progressive Complexity

  1. Self-evaluating expressions: Numbers, strings, booleans
  2. Variable lookup: Simple bindings
  3. Conditionals: if expressions
  4. Lambda and application: Basic function calls
  5. Let expressions: Local bindings
  6. Closures: Functions that capture environment
  7. Metacircular test: eval.scm interpreting simple expressions
  8. Full metacircular: eval.scm interpreting itself

Test File Organization

Create separate test files for different concerns:

  • test-env.scm - Environment operations
  • test-closure.scm - Closure creation and application
  • test-eval-basic.scm - Basic evaluation
  • test-metacircular.scm - Self-interpretation

Common Pitfalls

Shell Command Issues

When running tests via shell commands:

  • echo -e behavior varies between shells; prefer printf for consistent escape handling
  • Always verify test output format before debugging code issues

Premature Workarounds

Avoid these until root cause is understood:

  • Renaming primitives to avoid "collisions"
  • Transforming constructs (let to lambda) without understanding why direct implementation fails
  • Adding special cases for specific test files

Incomplete Understanding of Levels

Before implementing fixes, clearly answer:

  • At which level does evaluation happen for each sub-expression?
  • Which environment is used at each point?
  • How are closures represented when they cross interpretation boundaries?

Over-reliance on Debug Mode

When using debug/trace output:

  • Analyze the output systematically rather than skimming
  • Look for patterns in where success and failure diverge
  • Create hypotheses and test them explicitly

Verification Checklist

Before considering the implementation complete:

  • All self-evaluating expressions work at levels 1 and 2
  • Variable lookup works with shadowing at all levels
  • Lambda creates closures that capture correct environment
  • Closure application extends the closure's environment, not the call site's
  • Let expressions work for nested lets at all levels
  • Recursive functions work (requires proper environment setup)
  • The evaluator can evaluate itself running simple programs
  • The evaluator can evaluate itself running programs with closures

平台分布

Claude Code

25.14%
按下载量换算66

Gemini CLI

24.12%
按下载量换算64

Antigravity

15.71%
按下载量换算41

windsurf

13.22%
按下载量换算35

OpenCode

6.93%
按下载量换算18

Codex

3.54%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/letta-ai/skills --skill schemelike-metacircular-eval;npx skills add letta-ai/skills --skill "schemelike-metacircular-eval" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。

继续浏览同类 Skills