Skip to content

oioio-space/unpixel / .claude/skills/research-grounding

research-grounding

Verified

Use when answering ANY non-trivial question or doing research in this project — ground the answer in current external sources (web state-of-the-art, GitHub prior art, official docs) instead of memory, THEN go beyond the existing to find improvements and out-of-the-box approaches. A UserPromptSubmit hook reinforces this on every prompt.

Research Grounding

Goal: answer from current reality, not stale memory — then improve on it. Before giving a non-trivial or technical answer — especially anything about libraries, tools, versions, APIs, best practices, or "how do people do X" — consult external sources, let them shape the answer, and then push past the existing: find improvements and out-of-the-box approaches rather than just replicating prior art.

The workflow

  1. State of the art (web) — WebSearch for recent (2025–2026) best practices and comparisons; WebFetch to read the authoritative page. Search before claiming "X is the standard."
  2. Prior art (GitHub) — use scripts/ghx.sh (= mise gh + rtk) to look at existing implementations, popular libraries, their licenses, activity, and how the original bishopfox/unredacter solved a problem. gh search repos/code, gh api.
  3. Official docs — the microsoft-docs MCP for Microsoft/Azure/.NET; claude-code-guide for Claude Code/SDK/API questions; vendor docs via WebFetch. Verify signatures/versions rather than guessing.
  4. Deep questions — for broad, multi-source questions, invoke the deep-research skill.
  5. Go beyond the existing (improve & innovate) — once you understand the state of the art, don't stop at replicating it:
    • Critique it — what are its limitations, costs, or outdated assumptions?
    • Improve — is there a newer/faster/simpler/cheaper approach, a better library, or a way to combine ideas? Look for what the common solution misses.
    • Out-of-the-box — consider novel/unconventional approaches (a different algorithm, a different data representation, a tool from an adjacent domain) that could surpass prior art.
    • Recommend the best approach for the goal — even if it isn't the most common — and say why it beats the standard one.

How to apply

  • Prefer up-to-date primary sources over training memory; if memory and a current source disagree, trust the source and say so.
  • Compare ≥2 alternatives for any tool/library/approach choice, with the trade-off and a recommendation — and ask whether an improved or out-of-the-box option beats them all.
  • Cite what you used (URLs / repos) so the answer is verifiable.
  • Pin versions and check release dates — the ecosystem moves fast.
  • Calibrate: skip the research for trivial or mechanical asks (rename, format, "run the tests"); spend it where a wrong/stale answer would cost real time.

Notes

  • This is how every recent decision here was made (gitleaks vs trufflehog, govulncheck/gosec, syft/grype, mise backends, GPL-3.0 license check) — confirm with sources, then act.
  • Tools available: WebSearch, WebFetch, scripts/ghx.sh/gh, the microsoft-docs MCP, claude-code-guide, and the deep-research skill.
rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.Stars: 197kUpdated Claude CodeCodexVerifiedInstall
claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).Stars: 197kUpdated Claude CodeCodexUnverifiedInstall
doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.Stars: 197kUpdated Claude CodeCodexVerifiedInstall