How do I install the Search Governance skill?

Run npx skillmds@latest add shawnclybor/search-governance in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Search Governance safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Search Governance?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Search Governance free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Search Governance?

shawnclybor (@shawnclybor) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Search Governance by shawnclybor — Agent Skill

Name: Search Governance
Author: shawnclybor

Search Governance

Governance layer for all web search operations in the life-crm workflow. Enforces the mandatory use of Brave Search (`brave_web_search` via MCP) and blocks the built-in `WebSearch` tool. Use this skill whenever you are about to perform a web search, look something up online, search for documentation, find current information, or any request that involves querying the web — even if the user just says "look up", "search for", "find out", "google", or "what is the latest on." Also use when you instinctively reach for WebSearch or WebFetch — this skill redirects you to the correct tool. If in doubt about whether a web lookup is involved, use this skill — it's cheap to consult and expensive to use the wrong search tool.

shawnclybor Updated Sep 21, 2026

File contents

Search Governance Rules

These rules apply to ALL web search operations in the life-crm workspace.

The Rule

ALWAYS use brave_web_search (Brave Search MCP) for web lookups. NEVER use the built-in WebSearch tool.

Brave Search is configured as a local MCP server and is the only approved search tool for this project. The built-in WebSearch tool is prohibited because it bypasses the MCP layer and its results are less controllable.

If brave_web_search fails, debug the MCP connection — do not fall back to WebSearch.

If no Brave tool is present in the session at all, that is a blocker, not a licence to fall back. A blocked tool is not a fallback. Instead:

Do not use WebSearch.
Record the gap as a finding — write NOT PROBED — no approved web search tool in this session next to whatever the search would have established, and name what would settle it.
Say so in chat. A skill that silently cannot do its job is worse than one that refuses out loud.

Confirming a server is connected is not the same as exercising the search path. Where a connection has been verified but no live query run, say so, and treat a first failure as a live-path problem rather than as absence of the tool.

Tool Reference

Tool	Status	Use When
`brave_web_search`	Approved	All web lookups — docs, current events, research, troubleshooting
`brave_local_search`	Approved	Location-based searches (businesses, places)
`WebSearch`	Blocked	Never. Not even as a fallback.
`WebFetch`	Allowed	Fetching a specific URL you already have (not searching)

Brave Search Quick Reference

brave_web_search(
  query="your search query",   # max 400 chars, 50 words
  count=10,                     # results per page (1-20, default 10)
  offset=0                      # pagination (max 9, default 0)
)

Tips for effective queries:

Be specific: "Notion MCP query-data-source filter syntax" beats "Notion API"
Use site scoping: "site:developers.notion.com query data source"
For docs: include the product version or date range if relevant
For troubleshooting: include the exact error message in quotes

If Brave returns 0 results: don't fall back to WebSearch. Reformulate the query instead — drop site scoping, loosen exact-match quotes, try synonyms, or break a compound query into parts. If the topic is genuinely obscure and still returns nothing after 2–3 reformulations, report that to Shawn rather than guessing from memory.

When WebFetch Is Fine

WebFetch fetches a specific URL — it's not a search engine. It's fine for:

Reading a documentation page you already have the URL for
Fetching API references, GitHub READMEs, blog posts
Downloading content from a known URL

The governance concern is about searching (discovering new URLs), not fetching (reading a URL you already have).

Pre-Flight Checklist

Before any web search:

Am I using brave_web_search? (not WebSearch)
Is my query specific enough to get useful results?
If Brave fails, am I debugging the MCP connection instead of falling back?

---name: search-governancedescription: Governance layer for all web search operations in the life-crm workflow. Enforces the mandatory use of Brave Search (`brave_web_search` via MCP) and blocks the built-in `WebSearch` tool. Use this skill whenever you are about to perform a web search, look something up online, search for documentation, find current information, or any request that involves querying the web — even if the user just says "look up", "search for", "find out", "google", or "what is the latest on." Also use when you instinctively reach for WebSearch or WebFetch — this skill redirects you to the correct tool. If in doubt about whether a web lookup is involved, use this skill — it's cheap to consult and expensive to use the wrong search tool.---# Search Governance RulesThese rules apply to ALL web search operations in the life-crm workspace.## The Rule> **ALWAYS use `brave_web_search` (Brave Search MCP) for web lookups.> NEVER use the built-in `WebSearch` tool.**Brave Search is configured as a local MCP server and is the only approvedsearch tool for this project. The built-in `WebSearch` tool is prohibitedbecause it bypasses the MCP layer and its results are less controllable.If `brave_web_search` fails, debug the MCP connection — do not fall backto `WebSearch`.**If no Brave tool is present in the session at all**, that is a blocker, not alicence to fall back. A blocked tool is not a fallback. Instead:1. **Do not use `WebSearch`.**2. **Record the gap as a finding** — write **`NOT PROBED — no approved web search tool in this session`** next to whatever the search would have established, and name what would settle it.3. **Say so in chat.** A skill that silently cannot do its job is worse than one that refuses out loud.Confirming a server is connected is not the same as exercising the search path.Where a connection has been verified but no live query run, say so, and treat afirst failure as a live-path problem rather than as absence of the tool.## Tool Reference| Tool | Status | Use When ||------|--------|----------|| `brave_web_search` | **Approved** | All web lookups — docs, current events, research, troubleshooting || `brave_local_search` | **Approved** | Location-based searches (businesses, places) || `WebSearch` | **Blocked** | Never. Not even as a fallback. || `WebFetch` | **Allowed** | Fetching a specific URL you already have (not searching) |## Brave Search Quick Reference```brave_web_search( query="your search query", # max 400 chars, 50 words count=10, # results per page (1-20, default 10) offset=0 # pagination (max 9, default 0))```**Tips for effective queries:**- Be specific: `"Notion MCP query-data-source filter syntax"` beats `"Notion API"`- Use site scoping: `"site:developers.notion.com query data source"`- For docs: include the product version or date range if relevant- For troubleshooting: include the exact error message in quotes**If Brave returns 0 results:** don't fall back to `WebSearch`. Reformulate the queryinstead — drop site scoping, loosen exact-match quotes, try synonyms, or break acompound query into parts. If the topic is genuinely obscure and still returns nothingafter 2–3 reformulations, report that to Shawn rather than guessing from memory.## When WebFetch Is Fine`WebFetch` fetches a specific URL — it's not a search engine. It's fine for:- Reading a documentation page you already have the URL for- Fetching API references, GitHub READMEs, blog posts- Downloading content from a known URLThe governance concern is about *searching* (discovering new URLs), not*fetching* (reading a URL you already have).## Pre-Flight ChecklistBefore any web search:1. Am I using `brave_web_search`? (not `WebSearch`)2. Is my query specific enough to get useful results?3. If Brave fails, am I debugging the MCP connection instead of falling back?

Frequently asked questions