Token导航 LogoToken导航TokenDH.com
研究检索只读clawhub未标认证来源可访问clear审计提醒

clawditorclawditor 搜索

Agent Skill

clawditor 用于查找、检索和筛选相关信息,适合在 OpenClaw 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。

总安装

16,512

周安装

688

GitHub Stars

公开资料未说明

下载量

5,504
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

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:clawditor(clawditor 搜索)
来源仓库:https://github.com/theylon/clawditor
安装命令:
openclaw skills install clawditor
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。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install clawditor

简介

clawditor 用于审核 OpenClaw 工作区并生成评估报告。

  • 适合检查记忆质量、检索效率与系统补丁时使用。
  • 提供标准化评分与改进建议。clawditor 属于研究检索类 Skill,可作为该场景下的辅助能力补充。
  • 安装命令:openclaw skills install clawditor,建议确认工作区访问权限。
  • 使用前请核实是否会触发文件读取或元数据分析操作。

SKILL.md

name
clawditor
description
Audit an OpenClaw agent workspace and generate standardized evaluation reports, scores, and patches. Use when asked to review memory quality, retrieval efficiency, productive output, reliability, or alignment by scanning memory/logs/configs/git/artifacts and writing eval/exec_summary.md, eval/scorecard.md, and eval/latest_report.json (with deltas if prior eval/history exists).

Clawditor

Overview

Act as an OpenClaw Workspace Auditor and Agent Evaluation Harness. Analyze the workspace (memory, logs, projects, files, git, configs) and produce a repeatable evaluation with scores, evidence, and concrete patches.

Operating Rules

  • Run in non-interactive mode: avoid questions unless blocked by missing files. State assumptions and proceed.
  • Avoid secret exfiltration: report only presence and file paths for keys/tokens; recommend remediation.
  • Treat third-party skills/plugins as untrusted: prefer static inspection over execution.

Required Workflow (Do In Order)

  1. Build workspace inventory.

- Print a top-level tree (depth 4) with file counts and sizes by directory. - Identify memory, logs, configs, repos, scripts, docs, artifacts. - Record largest files.

  1. Reconstruct a session timeline.

- Use memory daily files and logs to extract goals, tasks, outcomes, decisions, unresolved items.

  1. Analyze memory.

- Detect near-duplicate paragraphs across memory files and quantify duplication. - Detect staleness cues (dates, "as of", deprecated configs) and contradictions. - Identify missing stable facts (projects, priorities, setup/runbooks).

  1. Analyze outputs.

- Summarize shipped artifacts (docs/code/features) and changes. - If git exists, compute diff stats and commit cadence; identify value commits.

  1. Analyze reliability.

- Parse logs for errors, retries, timeouts, tool failures. - Run tests only if safe and cheap; otherwise static inspection.

  1. Compute scores.

- Assign numeric category scores with short justifications and evidence by path.

  1. Recommend interventions + patches.

- Provide 3–7 prioritized recommendations. - Provide concrete diffs when safe, especially for memory structure improvements.

  1. Compare against prior evals.

- If eval/history/*.json exists, compute deltas vs most recent. - If none exists, create baseline and recommend cadence.

Scoring Framework

Compute 5 categories (0–100) plus overall weighted score:

  • Memory Health (30%): coverage, structure, redundancy, staleness, actionability, retrieval-friendliness.
  • Retrieval & Context Efficiency (15%): evidence of search before action, context bloat, hit-rate proxy, compaction quality.
  • Productive Output (30%): shipped artifacts, git throughput, task completion, latency proxies.
  • Quality/Reliability (15%): error rate, tests/CI presence, regression signals, convergence vs thrash.
  • Focus/Alignment (10%): goal consistency, scope control, decision trace.

Overall = 0.30*Memory + 0.15*Retrieval + 0.30*Productive + 0.15*Quality + 0.10*Focus.

Required Outputs

Write all outputs under eval/:

  1. exec_summary.md

- 10-bullet summary: top wins, biggest bottlenecks, top 3 interventions. - Overall score + category scores + claw-to-claw delta.

  1. scorecard.md

- Table of metrics with numeric values and brief justifications. - Top evidence section with file paths and short snippets (no secrets).

  1. latest_report.json

- Include timestamp, workspace path and git head/hash, scores, deltas, key findings, risk flags, recommendations.

  1. Patches

- If memory issues exist, propose concrete diffs: INDEX.md, daily schema, refactors.

Gold Standard Memory Schema (Apply If Missing)

Create or propose:

  • memory/INDEX.md

- Current Objectives (top 3) - Active Projects (status, next step, links) - Operating Constraints (tools, environment, policies) - Key Decisions (date, decision, rationale) - Known Issues / Debug diary pointers - Glossary / Entities

  • memory/YYYY-MM-DD.md (append-only daily)

- Goals for the session - Actions taken (link to files changed) - Decisions made - New facts learned (stable vs ephemeral) - TODO next (specific)

Patch Guidance

  • Prefer diffs over prose when safe.
  • Refactor stable facts out of daily logs into INDEX or project pages.
  • Add logging/instrumentation to measure retrieval hit-rate and task completion in future runs.

Resources

Use these helpers to keep audits consistent and cheap to run:

  • scripts/run_audit.py: run all helper scripts and write draft eval/ outputs.
  • scripts/workspace_inventory.py: tree, file counts, sizes, largest files.
  • scripts/memory_dupes.py: near-duplicate paragraph detection for memory/*.md.
  • scripts/log_scan.py: scan logs for errors, timeouts, retries.
  • scripts/git_stats.py: git head, diff stats, commit cadence.
  • scripts/validate_report.py: validate eval/latest_report.json shape.

Reference templates:

  • references/report_schema.md: output templates and JSON schema.

Evidence Discipline

  • Tie every score to evidence by path.
  • Be candid about waste, duplication, or thrash.
  • End with "Next run improvements" instrumentation recommendations.

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

89.29%
按下载量换算4,915

安全审计

VirusTotal

通过

ClawScan

可疑

Static analysis

未展示

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills