Anti-hallucination toggle for Claude Code. Activates three constraints from Anthropic's documentation that force Claude to cite sources, say "I don't know" when unsure, and ground responses in direct quotes.
As a plugin (recommended):
/plugin marketplace add franko14/research-mode
/plugin install research-mode@franko14-research-mode
As a skill:
Clone the repo into your .claude/skills/ directory or add it via the Claude Code skills UI.
/research-mode:research
Or with a topic:
/research-mode:research what caused the Change Healthcare breach
Say "exit research mode" to turn it off.
Three constraints activate simultaneously:
- Say "I don't know" - no guessing, no inferring. If there's no credible source, Claude says so.
- Cite everything - every claim must reference a file, URL, paper, or named source. Unsourced claims get retracted.
- Quote first, then analyze - responses are grounded in word-for-word quotes from source material, not paraphrased summaries.
- Not always-on. It's a toggle. Turn it on for research, off for creative work.
- Not slow. Claude still uses tools in parallel and works efficiently.
- Not restrictive on new ideas. You can synthesize across sources, but inputs must be grounded.
LLMs hallucinate. When you're doing research that matters, you need guardrails that force citation discipline. This plugin packages Anthropic's own recommendations into a one-command toggle.
I built this while running GTM, investor outreach, and content ops for KTLYST entirely through Claude Code. When your AI assistant is writing your pitch decks, researching competitors, and drafting investor briefs, hallucinated facts aren't a minor annoyance. They're a credibility risk. This toggle exists because I needed it.
I also consult with founders on AI-native operations, from Claude Code workflows to agent pipelines. If you're running your startup through AI and want to talk about what's working, reach out at assaf@ktlystlabs.com.