Research Mode
Turn on strict citation and source-grounding behavior before market, technical, or competitive research so agents say “I don’t know” instead of inventing facts.
Overview
research-mode is a journey-wide agent skill that enforces citation-minded, source-grounded answers and honest “I don’t know” responses—usable whenever a solo builder needs verified facts before committing.
Install
npx skills add https://github.com/assafkip/research-mode --skill SKILL.mdWhat is this skill?
- Toggleable research mode: enforce citations and source grounding when on, normal creative output when off
- Explicit “I don’t know” behavior instead of fabricating URLs, stats, or API details
- Pairs with any agent session where factual accuracy matters more than brainstorming flair
- Lightweight procedural guardrail—no external APIs required
- Reusable across competitive scans, compliance checks, and pre-build technical lookups
Adoption & trust: 138 GitHub stars.
What problem does it solve?
Your agent confidently states competitor pricing, API limits, or legal claims you cannot trace to a real source.
Who is it for?
Solo builders running competitive research, tool comparisons, or pre-ship fact checks in the same repo where the agent also drafts code and copy.
Skip if: Pure brainstorming, fiction, or tasks where you already have an approved spec and only want fast implementation without citation overhead.
When should I use this skill?
Before factual research, competitive analysis, or any task where invented citations would create downstream risk; toggle off for creative writing and open-ended ideation.
What do I get? / Deliverables
Research sessions produce traceable reasoning, explicit uncertainty, and fewer invented citations until you toggle back to creative mode.
- Source-grounded research answers with explicit uncertainty
- Reduced fabricated citations during research sessions
Recommended Skills
Journey fit
Useful at every journey phase - explore requirements and options before committing to a direction.
Where it fits
Compare three SaaS pricing pages without inventing tier names the agent never saw.
Check whether a regulation actually applies before locking MVP scope.
Read unfamiliar REST docs and flag what is documented versus assumed.
Draft AI-search FAQs only from sources you can name in the answer.
Interpret benchmark reports and separate cited metrics from guesses.
How it compares
Use as a session-wide behavior guardrail instead of repeating “cite your sources” in every chat message.
Common Questions / FAQ
Who is research-mode for?
Indie and solo builders who rely on coding agents for market and technical research and need a simple on/off switch for hallucination-resistant behavior.
When should I use research-mode?
Use it during Idea competitor and audience research, Validate scoping when checking claims, Build integrations when reading unfamiliar APIs, Launch SEO/geo fact gathering, and Grow analytics when interpreting metrics you did not measure yourself.
Is research-mode safe to install?
It is a procedural SKILL.md behavior pattern with no bundled credentials; review the Security Audits panel on this Prism page before enabling it in production workflows.
SKILL.md
READMESKILL.md - Research Mode
Anti-hallucination research mode. Toggle on to enforce citation requirements, source grounding, and "I don't know" behavior. Toggle off for creative work. # research-mode { "name": "research-mode", "description": "Anti-hallucination research mode. Toggle on to enforce citation requirements, source grounding, and \"I don't know\" behavior. Toggle off for creative work." }