snowflake-ai-research/arcticswarm/tool-usage-policy-browsing-corpus
Claude Code and agent skill by snowflake-ai-research/arcticswarm
- community
Checking that the SKILL.md file is still in the repository.
Tool usage guidelines for corpus research agents. Covers parallel tool calling, search strategy, and the search-then-fetch workflow for the static document corpus.
Install
npx vskill@latest install snowflake-ai-research/arcticswarm/tool-usage-policy-browsing-corpusSKILL.md
The same SKILL.md was published earlier as majiayu000/claude-skill-registry/tool-usage-policy-browsing-corpus.
Tool Usage Policy (Corpus Browsing)
General
- Call multiple tools in a single response when they are independent.
- Prefer specialized tools over shell commands:
read_fileinstead of cat/head/tail
Corpus Research Workflow
Use these tools for effective corpus research:
- `web_search` — Search the document corpus for relevant text chunks.
Returns scored snippets (~1200-1400 chars each). Use this to discover relevant information and assess what documents to investigate further.
- `web_fetch` — Retrieve the full text of a document from the corpus.
Pass a descriptive search query (NOT a URL) describing what document you want to read. Returns the complete document text (~2500-7200 chars). Use when search snippets aren't sufficient and you need the full context.
- `calculator` — Evaluate numeric expressions. NEVER calculate mentally.
Search-Then-Fetch Principle
- Use
web_searchfirst to find relevant chunks and identify promising documents. - When you need more context than the snippets provide, use
web_fetchwith
a targeted query to retrieve the full document text.
web_fetchsearches the corpus independently — it does NOT take URLs.
Describe what you want to read in the query.
Search Strategy
- Start with broad queries to understand the landscape, then narrow down.
- Use multiple search queries when the topic has different facets or
terminology — do not rely on a single query.
- If the first search returns no useful results, rephrase the query using
synonyms or alternative keywords before giving up.
Source Attribution
- Include evidence from the corpus for every factual claim.
- When multiple sources conflict, note the discrepancy and cite all sides.
Do Not Guess
- Do NOT claim a fact without supporting evidence from the corpus — if you
cannot find evidence, say so explicitly.
- Cross-check key numbers or dates across multiple document chunks when possible.
Sharing Results (Swarm Mode)
- Post search findings to
#key-findingsor#discoveries. - Include a brief summary of what you found and why it's relevant.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingarcticswarm/skills/tool-usage-policy-browsing-corpus/SKILL.md
Popularity
- Repository stars
- 13
- 7-day trend score
- 18
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/snowflake-ai-research/arcticswarm/tool-usage-policy-browsing-corpus)