quick-research

Automatically perform a fast, reliable web lookup when the user conversationally says "araştırır mısın", "bir bak", "bakabilir misin", "güncel mi", "fiyatı ne", "hangisi", "doğru mu", or asks for a sm

SKILL.md

GitHub

Quick Research

Answer the user's actual question quickly without sacrificing source integrity. This skill owns ordinary lookups; it is not a shortened substitute for a high-stakes or broad investigation.

Workflow

  1. Define the lookup. Resolve exact entity/product/version/location/date and the one or two facts or comparison criteria that matter. Ask only if ambiguity would change the answer.
  2. Check freshness. Identify which claims may have changed and use live sources for them. Never fill a current fact from model memory.
  3. Choose sources by claim. Prefer the official/primary owner for status, rule, specification, price, schedule, or announcement. Add a strong independent source when interpretation, performance, controversy, or bias matters.
  4. Open the evidence. Inspect the actual page or record; do not rely on a search snippet, headline, generated summary, or result ranking.
  5. Check incentives and independence. Label seller, issuer, sponsored, affiliate, advertorial, or community sources. Several pages copying one origin count as one source.
  6. Resolve conflict. Compare date, scope, version, geography, units, definitions, and source authority. If still unresolved, show the disagreement rather than averaging or guessing.
  7. Verify the draft. Check names, dates, numbers, units, links, and whether every citation supports the nearby claim. Use $evidence-integrity-guard when available for disputed, commercial, time-sensitive, or consequential facts.

Depth boundary

Switch to $deep-research when the request needs broad coverage, several research questions, due diligence, an exhaustive scan, an evidence table, deliberate counterevidence, or a consequential recommendation. Use a domain skill for medicine, finance, law, security, or another specialist decision. A simple user phrasing controls answer length, not the required safety depth.

Output

Lead with a direct answer or winner. Then give the 2–5 facts that drive it, the relevant date/location/version, a concise caveat or conflict if any, and citations beside the claims they support. Do not force a generic report template, pricing section, pros/cons list, or minimum source count when the question does not need it.

Rules

  • Never invent or infer a source, quotation, statistic, date, price, feature, or current status.
  • Never present advertising, seller copy, an affiliate ranking, or an issuer claim as independent validation.
  • Never say “all sources” or “the whole market” was checked when only a small lookup was performed.
  • State exactly what could not be verified or accessed.
  • Keep the result compact and actionable unless the user asks for detail.