search-everything

Comprehensive multi-source research on any topic. Use when the user says "search everything about X", "find all issues about X", "what exists for X", or before filing any issue/PR to ensure no dupes exist. Fans out to GitHub, web, knowledge/, npm/registry, and URL-reads in parallel.

Search Everything

Fan-out research across every available source. Run ALL searches in parallel. Deduplicate. Report findings with citations.

Step 1: Parse the query

Extract from the user's message:

• Topic: what to search (e.g. "lyrics translation", "SMTC Windows")
• Repos: any GitHub repos mentioned or implied (e.g. if user is talking about SimpMusic → maxrave-dev/SimpMusic)
• Context: is this for filing an issue? checking if feature exists? researching an API?

Step 2: Fan out (ALL in parallel)

2a. GitHub — issues + PRs

For each relevant repo, run simultaneously:

gh issue list --repo <owner>/<repo> --search "<topic>" --state all --limit 20 --json number,title,state,labels
gh pr list --repo <owner>/<repo> --search "<topic>" --state all --limit 10 --json number,title,state

Also search across all of GitHub:

gh search issues "<topic>" --limit 20 --json repository,title,number,state,url
gh search repos "<topic>" --limit 10 --json fullName,description,stargazerCount,url

2b. Web search (minimum 2 independent queries)

Query 1 — general: mcp__chirag127__keenable-web-search_search_web_pages Query 2 — specific/technical: mcp__chirag127__oevortex-ddg-search_web-search Query 3 (if needed) — news/recent: mcp__chirag127__linkupplatform-linkup-mcp-server_linkup-search

Use different phrasings. Never rely on a single web search.

2c. Workspace knowledge

python C:/D/oriz/scripts/okf-prompt-lookup.py "<topic>" --limit 5

Read each returned file.

2d. Package registries (if topic is a library/tool)

npm view <package> description homepage repository

Or check PyPI, pub.dev, crates.io as appropriate.

2e. URL reads

For any specific URL surfaced in 2a-2d that looks authoritative, fetch with mcp__fetch__fetch.

Step 3: Deduplicate + synthesise

Group findings by:

• Already exists (closed issue, merged PR, shipped feature) → cite link
• Open / in-progress (open issue, open PR) → cite link
• Not found → confirm absence across all sources searched

Step 4: Report

## Search: <topic>

### Already exists
- [<title>](<url>) — <one-line summary>

### Open / in-progress  
- [<title>](<url>) — <one-line summary>

### Not found
- <what was searched, what came back empty>

### Sources checked
- GitHub: <repos searched>
- Web: <queries run>
- Knowledge: <files matched>

Rules

• Minimum 2 web searches with different phrasings. One search = insufficient.
• Always check knowledge/ — workspace decisions may already answer the question.
• Cite every claim with a URL or file path. No assertions without sources.
• Parallel where possible — don't serialise if sources are independent.
• Short report — bullets only. No prose paragraphs.
• No hallucinations — if you didn't find it, say "not found in sources searched", not "it probably doesn't exist".

Trigger phrases

• "search everything about X"
• "find all issues about X"
• "what exists for X"
• "is there already a X"
• "check if X is filed"
• "research X before filing"