All templates
⚡ Advanced Techniques·Works on: claude, chatgpt
Constitutional Self-Critique
Technique: Constitutional AI
Define principles up front, generate the answer, then audit it against each principle. Anthropic-style safety.
Advanced#constitutional#self-critique#safety
0/2
Fill the template
2 placeholders left.
~186tokens
Live preview· click any pill to jump to its field
Use Constitutional Self-Critique. What the prompt produces: Constitution (the principles every output must satisfy): Protocol: **Stage 1 — Draft.** Produce the output. Don't pre-hedge for the critique step. **Stage 2 — Critique.** For each numbered principle, do an explicit check: - Principle N: PASS / FAIL - If FAIL: quote the exact phrase that violates it. State which principle and why. Do not soften. "Mostly fine" is not a valid verdict. **Stage 3 — Revise.** If any FAIL, rewrite to fix ONLY the flagged violations. Don't change anything that passed. Return the final revised output, then a summary changelog: which principles failed in the original draft and what changed.
Inputs0 of 2
›See the lazy version this template replaces
Before — the lazy prompt
<task>
Why it works
- Explicit numbered principles convert vague 'be good' into auditable rules.
- Per-principle PASS/FAIL with quoted evidence prevents soft-pass output.
- Strict 'only fix flagged things' rule prevents over-revising into a different draft.
- Changelog at the end is the audit trail for downstream review.
Make this one yours
Replace the bracketed placeholders, then paste into the Prompt Fixer to lint your customisation before hitting send.