wiki-manager

管理基于单一JSON索引的wiki关键词池，支持核心、缓冲、新词三池循环，自动升降级及淘汰机制，保持索引稳定不膨胀。

Security Scan

The skill's purpose (manage a single JSON wiki index) matches its instructions, but there are a few coherence gaps — notably undeclared tool dependencies (jq) and instructions that allow reading arbitrary source paths — so review before installing.

Purpose & Capability

Name/description align with the provided instructions: a single wiki.json and three-pool lifecycle. However SKILL.md relies on the jq command (and standard shell tools like cat/cp) while the skill metadata declares no required binaries — this is an undeclared dependency that should be declared or removed.

Instruction Scope

Instructions use exec to run shell commands (cat, jq, cp) and explicitly tell the agent to read files referenced in entry.source (e.g., memory/..., skills/...). That is coherent for restoring detailed content, but it permits reading arbitrary paths referenced in wiki.json and thus could access other skills' files or other local data unless those source paths are constrained. No guidance or path whitelisting is provided.

Install Mechanism

This is instruction-only (no install spec) which is low-risk. Note: because the runtime commands require jq, the lack of an install specification or declared dependency means installation/runtime may fail or silently rely on an environment having jq — declare jq or provide an install step.

Credentials

The skill requests no environment variables, no credentials, and no config paths. That is proportional to its stated purpose.

Persistence & Privilege

always is false and autonomous invocation is default. The skill does not request elevated persistence or to modify other skills' configs. It suggests periodic checks but does not force always-on behavior.

What to consider before installing

This skill appears to do what it says (manage wiki.json with three pools) but take three precautions before installing: 1) Ensure the runtime environment has jq (or update the skill metadata to require/install it); otherwise the provided shell commands will fail. 2) Inspect any existing wiki.json entries and agree what paths are allowed in the source field — the skill's instructions will read files at those paths (e.g., memory/..., skills/...), which could expose unrelated local data if entries point to sensitive locations. 3) If you plan to let the agent run autonomously, consider limiting file-read permissions or adding explicit whitelists for source paths and add tests to prevent accidental exfiltration of secrets. If the authors can declare jq as a required binary and document/limit which source paths are valid, the remaining concerns would be resolved.

wiki-manager / Wiki 索引管理

版本：v1.0 创建：2026-04-08 更新：2026-04-08（空空建议，改为单 JSON 索引方案） 依据：空空方案 + 小蜂自想（两者高度一致） 铁三角：彧哥 + 空空 + 小蜂

---

核心理念

用工具化索引文件代替散 .md：

• 1 个 wiki.json = 所有关键词索引
• 500 条封顶
• 三池循环（core / buffer / new）
• 不存大段正文，内容交给向量库/memory

---

文件结构

wiki/
├── wiki.json       ← 索引文件（三池合一，全部词条在这里）
├── wiki.json.bak   ← 上一次备份（每次写入前自动更新）
└── SKILL.md       ← 本文件，管理脚本

---

wiki.json 结构

{
  "version": "1.0",
  "name": "小蜂Wiki索引",
  "pools": {
    "core":   { "name": "核心热词池", "limit": 200 },
    "buffer": { "name": "次热缓冲池", "limit": 200 },
    "new":    { "name": "新晋新词池", "limit": 100 }
  },
  "entries": [
    {
      "id": "001",
      "keyword": "四部曲",
      "pool": "core",
      "definition": "想→说→做→看，沟通决策流程",
      "source": "skills/4steps-to-wisdom/SKILL.md",
      "score": 5,
      "lastUsed": "2026-04-08",
      "created": "2026-04-08",
      "sourceType": "skill"
    }
  ]
}

字段说明：

字段	必填	说明
id	✅	唯一ID，格式"001"～"500"
keyword	✅	关键词（去重校验）
pool	✅	所在池：core / buffer / new
definition	✅	一句话定义
source	✅	引用路径（memory/xxx 或 skills/xxx）
score	✅	使用频率评分（1-5）
lastUsed	✅	最后使用日期 YYYY-MM-DD
created	✅	创建日期 YYYY-MM-DD
sourceType	❌	来源类型：skill / memory / discuss

---

三池容量

池	上限	说明
core	200	最高频，核心专属
buffer	200	次高频，缓冲梯队
new	100	新增词，观察期
合计	500	永不膨胀

---

循环规则

new 高频用 → 升 buffer
buffer 高频用 → 升 core
core 长期不用 → 降 buffer
buffer 长期不用 → 降 new → 淘汰

降级/淘汰触发：最后使用距今 > 30 天

---

管理脚本（无工具版，用 exec）

查词（按池）

# 查 core 池
cat wiki/wiki.json | jq '.entries | map(select(.pool == "core"))'

# 模糊搜索
cat wiki/wiki.json | jq '.entries | map(select(.keyword | contains("四部")))'

查所有（高效加载）

cat wiki/wiki.json    # 一次读完所有词条，毫秒级

新增词条

# 编辑 wiki/wiki.json，手动追加 entries[]
# 注意：先检查是否已存在（keyword 去重）

升降级

# 将某词从 new 升到 buffer（score >= 3 时触发）
# 修改 wiki.json 中对应 entry 的 pool 字段

淘汰检查（每次心跳或定期）

# 检查 buffer/new 池中最后使用 > 30 天的词
cat wiki/wiki.json | jq '.entries | map(select(.lastUsed < "2026-03-09" and .pool != "core"))'

备份（写入前自动）

cp wiki/wiki.json wiki/wiki.json.bak

---

compact 恢复流程

compact 后认知丢失 → 执行以下步骤恢复：

1. cat wiki/wiki.json         # 加载所有词条（毫秒级）
2. 读取 definition 字段       # 恢复一句话认知
3. 如需详细内容 → 从 source 字段路径读取 memory/ 向量库

---

维护节奏

动作	频率
新词入 new 池	按需，随时
升降级检查	每天开机
淘汰检查	每周
备份	每次 wiki.json 写入前

---

注意事项

• 每次写入前先 cp wiki/wiki.json wiki/wiki.json.bak
• wiki.json 损坏时用 cp wiki/wiki.json.bak wiki/wiki.json 恢复
• compact 后 reload 一次 wiki.json 即可恢复全部词条认知
• 不要存大段正文，只存引用路径，详细内容在 memory 向量库

---

Wiki 索引管理 v1.0 - 轻量高效，铁三角共识