一键导入
knowledge-gate
Knowledge documentation gatekeeper. Dynamic domain expert analysis to find optimal placement and connections.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Knowledge documentation gatekeeper. Dynamic domain expert analysis to find optimal placement and connections.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Tycono usage guide and project understanding. WHEN: user asks "how does tycono work", "what can tycono do", "how to use the AI team", "what are the commands", "what is a wave", "what is an agency", "how to check results", "where are the logs", "how to see what the team did", first time using tycono. WHEN NOT: user is already running a wave, asking about unrelated code, general coding questions.
Browser automation for AI agents using agent-browser CLI. Token-efficient alternative to Playwright MCP with accessibility-first element selection.
AKB document hygiene checker. Finds documents that confuse AI agents.
| name | knowledge-gate |
| description | Knowledge documentation gatekeeper. Dynamic domain expert analysis to find optimal placement and connections. |
| allowed-tools | Read, Glob, Grep, Bash, Agent |
"Prevent mindless document creation. Domain experts find optimal connections."
| Pattern | Problem |
|---|---|
| New insight -> create new doc immediately | Duplicates existing docs |
| Similar topics scattered across files | Inefficient search, AI confusion |
| More docs != better knowledge | Defeats AKB purpose |
| Isolated docs with no links | Discovery failure |
When documenting new knowledge:
# Search with keywords
grep -rn "{keyword1}\|{keyword2}\|{keyword3}" --include="*.md" .
# Find similar topic docs
find . -name "*.md" | xargs grep -l "{core_topic}"
Hub pattern: {folder-name}.md as folder entry point, or akb_type: hub frontmatter.
# Find Hub documents
find . -name "*.md" -exec sh -c 'dir=$(dirname "$1"); base=$(basename "$dir"); fname=$(basename "$1" .md); [ "$base" = "$fname" ] && echo "$1"' _ {} \;
grep -rl "akb_type: hub" --include="*.md" .
For every discovered Hub, run an Expert Agent in parallel:
Each expert evaluates:
Relevance criteria (strict):
| Relevance | Criteria | Example |
|---|---|---|
| HIGH | Without this insight, my domain docs are incomplete | "DB migration design" -> Architecture Hub |
| MEDIUM | Nice to reference but not required | "DB migration" -> Knowledge Hub |
| NONE | No direct relation to my domain | "DB migration" -> Company Hub |
Anti-bias rule: Don't force connections. Only HIGH = direct dependency.
| Rule | Description |
|---|---|
| Only HIGH gets action | MEDIUM/NONE are informational only |
| Pick 1 primary Hub | If multiple HIGH, choose most direct |
| Max 3 connections | Limit total link work to 3 |
| Deduplicate hierarchy | If parent Hub is HIGH, skip child |
| Overlap | Threshold | Action |
|---|---|---|
| HIGH (70%+) | Core topic matches | Add section to existing doc |
| MEDIUM (30-70%) | Related but different angle | New doc + cross-link |
| LOW (<30%) | Independent topic | New doc (register in Hub) |
| None | 0 search results | New doc |
Run akb-linter on new/modified document.
| Overlap | Action |
|---|---|
| 70%+ | Add to existing document (preferred) |
| 30-70% | Reference existing + cross-link |
| <30% | New document (justify + register in Hub) |
"1 doc added = maintenance cost increased" "Strengthen existing > Create new" "Isolated doc = dead doc"