Token导航 LogoToken导航

Agent Skill · 开发

secretary(secretary 分析)

用于辅助安全审计、权限检查、凭据风险、认证流程和常见漏洞排查。它适合让 Agent 梳理敏感配置、检查依赖风险、分析鉴权逻辑或生成安全复核清单。使用时不能把工具输出直接当最终结论,涉及密钥、令牌、用户数据或生产系统时,应先确认最小权限、脱敏方式和操作边界。

secretary 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为权限未标注;许可证信息还不明确,第三方扫描有未通过项,按安装量在开发类的 21,746 个 Skill 里排第 14,913。

总安装 649周安装 26GitHub Stars 29下载 210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
unknown

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:secretary(secretary 分析)
来源仓库:https://github.com/mwguerra/claude-code-plugins
仓库路径:skills/secretary
安装命令:
npx skills add https://github.com/mwguerra/claude-code-plugins --skill secretary
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 npx skills 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。

skills.shnpx skills
npx skills add https://github.com/mwguerra/claude-code-plugins --skill secretary

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

用于辅助安全审计、权限检查和凭据风险分析。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。适合梳理敏感配置、检查依赖风险或生成安全复核清单。通过 npx 命令从指定仓库安装并使用该技能。不能将工具输出直接作为最终结论,需确认最小权限和操作边界。

secretary 属于开发类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/secretary

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/mwguerra/claude-code-plugins --skill secretary;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

    权限没有明确标注,第一次使用时只给一个无关紧要的任务,观察它到底读了哪些文件和网络。

常见问题

6 条

secretary 怎么安装?

在终端执行 npx skills add https://github.com/mwguerra/claude-code-plugins --skill secretary。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

secretary 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。

secretary 需要什么权限?

权限级别没有明确标注。第一次使用时建议只给无关紧要的任务,观察它实际读写了哪些内容。

secretary 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

secretary 用的人多吗?

公开数据是累计安装 649 次,近一周安装 26 次,下载 210 次,GitHub 29 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。

secretary 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

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。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
secretary

secretary 分析 · 本页

64929权限未标注未明确未通过

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

Secretary Skill

Capture commitments, record decisions, track ideas, manage sessions, and maintain the full knowledge base.

When to Use

  • Analyzing conversation for commitments, decisions, or ideas
  • Recording a new commitment, decision, or idea manually
  • Updating the status of a tracked item (complete, defer, cancel)
  • Managing queue items and worker processing
  • Updating the knowledge graph
  • Managing goals and milestones
  • Accessing or managing encrypted memory entries
  • Checking worker and queue status

Database Locations

# Main database
SECRETARY_DB_PATH="$HOME/.claude/secretary/secretary.db"

# Encrypted memory database
SECRETARY_MEMORY_DB_PATH="$HOME/.claude/secretary/memory.db"

# Configuration
SECRETARY_CONFIG_FILE="$HOME/.claude/secretary.json"

# Scripts
PLUGIN_ROOT="$HOME/.claude/plugins/secretary"  # or wherever installed

Commitment Management

Recording a Commitment

-- Get next ID
SELECT 'C-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, 3) AS INTEGER)), 0) + 1) as next_id
FROM commitments;

-- Insert
INSERT INTO commitments (
    id, title, description, source_type, source_session_id,
    source_context, project, assignee, stakeholder,
    due_date, due_type, priority, status
) VALUES (
    :id, :title, :description, :source_type, :session_id,
    :context, :project, :assignee, :stakeholder,
    :due_date, :due_type, :priority, 'pending'
);

Updating Commitments

-- Complete
UPDATE commitments SET
    status = 'completed', completed_at = datetime('now'), updated_at = datetime('now')
WHERE id = :id;

-- Defer
UPDATE commitments SET
    status = 'deferred', deferred_until = :date,
    deferred_count = deferred_count + 1, updated_at = datetime('now')
WHERE id = :id;

-- Cancel
UPDATE commitments SET
    status = 'canceled', updated_at = datetime('now')
WHERE id = :id;

-- Change priority
UPDATE commitments SET
    priority = :new_priority, updated_at = datetime('now')
WHERE id = :id;

Listing Commitments

-- All pending (sorted by priority)
SELECT id, title, due_date, priority, project, status
FROM commitments
WHERE status IN ('pending', 'in_progress')
ORDER BY
    CASE WHEN due_date < date('now') THEN 0 ELSE 1 END,
    CASE priority WHEN 'critical' THEN 1 WHEN 'high' THEN 2 WHEN 'medium' THEN 3 ELSE 4 END,
    due_date ASC;

-- By project
SELECT id, title, due_date, priority, status
FROM commitments
WHERE project = :project AND status IN ('pending', 'in_progress')
ORDER BY priority DESC;

-- Overdue only
SELECT id, title, due_date, priority, project
FROM commitments
WHERE status IN ('pending', 'in_progress') AND due_date < date('now')
ORDER BY due_date ASC;

Detection Patterns

Look for these phrases in conversation:

Commitments:

  • "I will...", "I'll...", "Let me..."
  • "We should...", "We need to..."
  • "TODO:", "FIXME:", "Follow up on..."
  • "Don't forget to...", "Make sure to..."
  • "Remind me to...", "Get back to..."

Decision Recording

Recording a Decision

-- Get next ID
SELECT 'D-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, 3) AS INTEGER)), 0) + 1) as next_id
FROM decisions;

-- Insert
INSERT INTO decisions (
    id, title, description, rationale, alternatives,
    consequences, category, scope, project,
    source_session_id, source_context, status, tags
) VALUES (
    :id, :title, :description, :rationale, :alternatives_json,
    :consequences, :category, :scope, :project,
    :session_id, :context, 'active', :tags_json
);

Updating Decisions

-- Supersede
UPDATE decisions SET
    status = 'superseded', superseded_by = :new_decision_id,
    updated_at = datetime('now')
WHERE id = :old_id;

-- Reverse
UPDATE decisions SET
    status = 'reversed', updated_at = datetime('now')
WHERE id = :id;

Detection Patterns

Decisions:

  • "Decided to...", "The decision is..."
  • "Let's go with...", "We'll use..."
  • "The approach is...", "The plan is..."
  • "From now on...", "Going forward..."
  • "Instead of...", "Rather than..."

Extraction Process

  1. Identify decision phrase
  2. Extract what was decided
  3. Look for rationale ("because", "since", "due to")
  4. Identify alternatives mentioned ("instead of", "rather than")
  5. Categorize: architecture, process, technology, design
  6. Determine scope: project-wide, feature, component

Idea Capture

Recording an Idea

-- Get next ID
SELECT 'I-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, 3) AS INTEGER)), 0) + 1) as next_id
FROM ideas;

-- Insert
INSERT INTO ideas (
    id, title, description, idea_type, category,
    project, source_session_id, source_context,
    priority, effort, potential_impact, status, tags
) VALUES (
    :id, :title, :description, :type, :category,
    :project, :session_id, :context,
    :priority, :effort, :impact, 'captured', :tags_json
);

Updating Ideas

-- Start exploring
UPDATE ideas SET status = 'exploring', updated_at = datetime('now') WHERE id = :id;

-- Park for later
UPDATE ideas SET status = 'parked', updated_at = datetime('now') WHERE id = :id;

-- Mark done
UPDATE ideas SET status = 'done', updated_at = datetime('now') WHERE id = :id;

-- Discard
UPDATE ideas SET status = 'discarded', updated_at = datetime('now') WHERE id = :id;

Goal Management

Creating a Goal

SELECT 'G-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, 3) AS INTEGER)), 0) + 1) as next_id
FROM goals;

INSERT INTO goals (
    id, title, description, goal_type, timeframe,
    parent_goal_id, project, target_value, target_unit,
    target_date, status, milestones, related_commitments
) VALUES (
    :id, :title, :description, :type, :timeframe,
    :parent_id, :project, :target_value, :target_unit,
    :target_date, 'active', :milestones_json, :related_json
);

Updating Goal Progress

UPDATE goals SET
    current_value = :value,
    progress_percentage = ROUND(100.0 * :value / NULLIF(target_value, 0), 1),
    updated_at = datetime('now')
WHERE id = :id;

Activity Timeline

Event Types

TypeDescription
session_startNew session began
session_endSession completed
commitmentCommitment extracted
commitment_completedCommitment marked done
decisionDecision recorded
goal_progressGoal updated
goal_completedGoal finished
commitGit commit made
external_changeChange detected from outside

Recording Activity

INSERT INTO activity_timeline (
    activity_type, entity_type, entity_id,
    project, title, details, session_id
) VALUES (:type, :entity_type, :entity_id, :project, :title, :details_json, :session_id);

Knowledge Graph

Node Types

TypeDescription
projectSoftware projects
technologyLanguages, frameworks, tools
personTeam members, stakeholders
conceptArchitectural patterns, methodologies
toolDevelopment tools, services

Creating/Updating Nodes

SELECT 'N-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, 3) AS INTEGER)), 0) + 1) as next_id
FROM knowledge_nodes;

INSERT INTO knowledge_nodes (id, name, node_type, description, properties, aliases)
VALUES (:id, :name, :type, :description, :properties_json, :aliases_json)
ON CONFLICT(id) DO UPDATE SET
    description = COALESCE(:description, description),
    interaction_count = interaction_count + 1,
    last_interaction = datetime('now'),
    updated_at = datetime('now');

Creating/Updating Edges

SELECT 'E-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, 3) AS INTEGER)), 0) + 1) as next_id
FROM knowledge_edges;

INSERT INTO knowledge_edges (id, source_node_id, target_node_id, relationship, strength, properties)
VALUES (:id, :source, :target, :relationship, :strength, :properties_json)
ON CONFLICT(id) DO UPDATE SET
    strength = MIN(strength + 0.1, 1.0),
    updated_at = datetime('now');

Relationship Types

  • uses - Project uses technology
  • knows - Person knows technology
  • owns - Person owns project
  • depends_on - Project depends on another
  • related_to - General relationship

Encrypted Memory

Manage sensitive data via the memory manager script:

SCRIPT="$PLUGIN_ROOT/scripts/memory-manager.sh"

# Add a memory entry
bash "$SCRIPT" add "AWS API Key" "AKIA..." api_key "my-project" "aws,production"

# Search memory
bash "$SCRIPT" search "API key"

# List by category
bash "$SCRIPT" list credential
bash "$SCRIPT" list api_key my-project

# View a specific entry
bash "$SCRIPT" show 1

# Delete an entry
bash "$SCRIPT" delete 1

# Check encryption status
bash "$SCRIPT" status

Categories: credential, api_key, ip_address, phone, secret, note, general

Queue Operations

Check Queue Status

SELECT status, COUNT(*) as count
FROM queue
GROUP BY status;

Force Process Queue

bash "$PLUGIN_ROOT/scripts/process-queue.sh" --limit 10

Check Worker State

SELECT last_run_at, last_success_at, last_error,
    items_processed, total_runs,
    last_vault_sync_at, last_github_refresh_at
FROM worker_state WHERE id = 1;

Daily Notes

-- Create or update daily note
INSERT INTO daily_notes (id, date)
VALUES (date('now'), date('now'))
ON CONFLICT(id) DO NOTHING;

-- Update with session data
UPDATE daily_notes SET
    sessions_count = sessions_count + 1,
    last_activity_at = datetime('now'),
    updated_at = datetime('now')
WHERE date = date('now');

ID Generation

# Pattern for all entity IDs
# C-0001 for commitments
# D-0001 for decisions
# I-0001 for ideas
# G-0001 for goals
# P-0001 for patterns
# N-0001 for knowledge nodes
# E-0001 for knowledge edges

# SQL pattern to get next ID:
SELECT '<PREFIX>-' || printf('%04d', COALESCE(MAX(CAST(SUBSTR(id, LENGTH('<PREFIX>') + 2) AS INTEGER)), 0) + 1)
FROM <table>;

Output Guidelines

When reporting captured items:

## Captured

### Commitment
- **ID:** C-0025
- **Title:** Implement caching layer
- **Priority:** High
- **Due:** This week
- **Source:** Conversation

### Decision
- **ID:** D-0018
- **Title:** Use Redis for caching
- **Category:** Architecture
- **Rationale:** Better performance for distributed systems

### Idea
- **ID:** I-0015
- **Title:** GraphQL migration
- **Type:** Exploration
- **Impact:** High

Silence Mode

When operating via hooks (capture.sh), work silently:

  • Do not output unless explicitly requested
  • Log to activity_timeline for later review
  • Items are reviewed via /secretary:status or /secretary:briefing

Error Handling

  • Skip malformed or ambiguous items during extraction
  • Log extraction failures to debug log
  • Continue processing on individual item failures
  • Retry failed queue items up to 3 times
  • Expire unprocessed queue items after 24 hours

Related Commands

  • /secretary:track - Manage commitments (add, complete, defer, list)
  • /secretary:decide - Record decisions with rationale
  • /secretary:idea - Capture ideas
  • /secretary:status - Show full dashboard
  • /secretary:briefing - Generate context briefing
  • /secretary:memory - Manage encrypted memory
  • /secretary:worker - Check/trigger worker processing

平台分布

Codex

35.47%
按下载量换算74

Claude

33.55%
按下载量换算70

Cursor

18.78%
按下载量换算39

Gemini CLI

9%
按下载量换算19

安全审计

Gen Agent Trust Hub

可疑

Socket

可疑

Snyk

未通过

权限和风险

权限未标注

当前来源未能明确判断权限范围,默认进入异常复核队列。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills