A skill for AI agents · think critically, ask better questions

Truth‑Seeking

Same scalpel. Every side.

Today’s world runs on divisive, heavily contested topics — and most of what you and your AI encounter has been shaped by whoever paid the most to shape it. Truth‑Seeking is a skill that teaches your agent to think critically and ask the right questions: weigh every side by the same standard of proof, chase claims to their source, and stay resilient to lopsided, well‑funded information campaigns.

⚡ Add it to your agent in one line
npx skills add rmtbb/truth-seeking
Works with Claude Code · Hermes · any agent that reads skills
01

The method, in two minutes

The whole discipline · sound onWorks on any charged topic
One rule makes the rest work: symmetry.

Whatever skepticism you apply to one side, apply equally to the other — including the side you instinctively favor. Credibility is earned per claim, never granted by category.

02

The six moves

⚖️

Apply symmetry

Same standard of proof for every side. Mainstream isn’t truth; “forbidden” isn’t truth. No category gets a free pass — and none gets auto‑discredited.

🗂️

Tier every claim

Documented — a record you can open. Alleged — a named source, no public proof. Open — unsettled. Never upgrade a claim because it’s loud or on “your” side.

🔗

Chase the source

“Sources say” and “fact‑checkers confirm” are pointers, not proof. Follow every claim to the primary record. Most confident claims die at the dead end.

↔️

Both directions

Criticizing a state, ideology, or institution isn’t bigotry toward a people — and it’s never a license for bigotry either. Don’t conflate the two; don’t weaponize the accusation.

🕳️

Spot the cage

“Anyone who disagrees is compromised.” “Anyone asking is a crank.” Both are traps. A claim that nothing could ever disprove isn’t strong — it’s unfalsifiable.

🧭

Follow the evidence

Name the manipulation on every side, ask who benefits, steelman before you judge — and land where the proof points, even when no side likes it.

unprovendisprovenproven  ·  most contested claims live in the unresolved middle

03

Install in one line

Adds a truth-seeking skill to your agent’s skills folder (~/.claude/skills/). Works with Claude Code, Hermes, and any agent that reads skill files. Then just say /truth-seeking — or let it trigger itself whenever a topic is polarized.

Easiest — skills CLI
npx skills add rmtbb/truth-seeking
Manual one‑liner
mkdir -p ~/.claude/skills/truth-seeking && \
curl -fsSL https://raw.githubusercontent.com/rmtbb/truth-seeking/main/SKILL.md \
  -o ~/.claude/skills/truth-seeking/SKILL.md
Clone
git clone https://github.com/rmtbb/truth-seeking.git ~/.claude/skills/truth-seeking
Copied to clipboard