Skip to content

hint-shu/deep-research / skills/quick-research

quick-research

Verified

Quick web research for simple questions — fast answer with 3-5 cited sources in ~1 minute. Use for fact-checks, simple lookups, "what is X", "latest version of Y", "who made Z". For anything deeper, use /research (L1) or /deep-research (L2).

Quick Research (L0)

Fast lightweight research for simple questions. No planner, no reflection, no multi-agent — just a quick multi-source search with a cited answer.

Position in ladder: L0 — the lightest tier. When in doubt between this and /research, prefer /research.

When to Use

  • Fact checks: "какая последняя версия Next.js", "когда вышел Bun 2.0"
  • Simple lookups: "что такое edge runtime", "кто автор библиотеки X"
  • Yes/no questions with supporting evidence
  • Any question the user frames as "быстро", "коротко", "кратко"

Do NOT use for:

  • Comparisons (use /deep-research)
  • Strategic decisions (use /expert-research)
  • Topics requiring depth (use /research or higher)

Budget

  • Time: ~1 min
  • Tavily credits: ~5
  • Sources read: 3–5
  • Output: 2–4 paragraphs with inline citations

Pipeline

v0.6.0+ fast path (Perplexity available):
  ASK — one Perplexity call → synthesized answer + citations (~5 sec)

Fallback path (Perplexity unavailable):
  1. SEARCH   — parallel Firecrawl + Tavily
  2. READ     — scrape top 3 results
  3. ANSWER   — short structured response with citations

Fast path: Perplexity (v0.6.0+)

If mcp__perplexity-ask__perplexity_ask is available AND DEEP_RESEARCH_DISABLE_PERPLEXITY is unset, try Perplexity first. It returns a synthesized answer with inline citations in one call — exactly L0's output shape, no scrape needed:

mcp__perplexity-ask__perplexity_ask with:
  messages: [
    {role: "user", content: "<user query>. Respond in 2-4 paragraphs with inline [N] citations followed by a Sources: list with [N] <URL> per line. Match the user's language (Russian/English)."}
  ]

Response is ready to forward to user. Typical latency: 5 seconds.

Fall back to the SEARCH→READ→ANSWER path below if:

  • Perplexity returns an error or rate limit
  • Response has fewer than 2 citations (quality check)
  • User explicitly wants independent verification (rare for L0)

Step 1: SEARCH (parallel multi-source)

Run these in PARALLEL in a single message:

firecrawl search "<query>" --limit 5 --json

And Tavily:

mcp__tavily__tavily_search with:
  query: <query>
  max_results: 5
  search_depth: "basic"

For Russian-language topics: add one extra search in Russian.

Step 2: READ (scrape top 3)

Pick the 3 most relevant/authoritative results and scrape them IN PARALLEL:

firecrawl scrape "<url1>" --only-main-content &
firecrawl scrape "<url2>" --only-main-content &
firecrawl scrape "<url3>" --only-main-content &
wait

Source priority:

  1. Official docs / primary sources
  2. Recent articles (2025–2026)
  3. Authoritative tech sources
  4. Skip: SEO spam, affiliate pages, paywalled content

Use head -80 and grep to find relevant sections. Don't read full files.

Step 3: ANSWER

Format the response as:

**[Direct answer in 1-2 sentences]**

[2-3 supporting paragraphs with inline citations [1], [2], [3]]

---
**Sources:**
1. [Title](URL)
2. [Title](URL)
3. [Title](URL)

Rules

  • Always cite — every factual claim needs [N] reference
  • Direct answer first — no preamble, no "let me research this"
  • Flag uncertainty — if sources disagree, say so briefly
  • Freshness — prefer 2025–2026 sources
  • Language — match user's language (Russian / English)
  • No artifacts on disk — this is a quick mode, don't save to .firecrawl/research/
  • No clarifying questions — if query is ambiguous, pick most likely interpretation and answer

Escalation

If during research you realize the question is more complex than expected (needs comparison, deep analysis, multiple perspectives), STOP and tell the user:

"Этот вопрос серьёзнее чем quick-research. Рекомендую /research (L1) или /deep-research (L2). Продолжить quick или переключиться?"

add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 114kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 114kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 114kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 114kUpdated Claude CodeCodexVerifiedInstall