constitutional-ai
Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers Claude's safety system.
- Score
- Unranked
- Evidence
- No signals yet
- Last commit
- Not known
- Listed
Install
armory install constitutional-ai --cli claudewrites the skill into.claude/skills/constitutional-ai/SKILL.mdListed as compatible
# fetches the source and writes it to:
.claude/skills/constitutional-ai/SKILL.mdNeeds the armory CLI · not on npm yet, build it from cli/ in the repository
What it is
Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers Claude's safety system.
When to use it
Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers Claude's safety system.
How to install / invoke
# Copy the skill into your .claude/skills/ directory
curl -sL https://raw.githubusercontent.com/davila7/claude-code-templates/main/cli-tool/components/skills/ai-research/safety-alignment-constitutional-ai/SKILL.md -o .claude/skills/safety-alignment-constitutional-ai/SKILL.md
Notes
Extracted from davila7/claude-code-templates, safety-alignment-constitutional-ai category.