affirmations — Claude Code skill: what it does, how to install
A skill for an agent stuck in a loop: it reads a short set of affirmations, re-checks what it actually knows, and takes one small step.
no license 242 installs ★ 31 updated 2026-08-31
npx skills add 101-skills/superpowers --skill affirmations -a claude-codeWhat affirmations does
The skill is a markdown file the agent opens on itself once the signs show up: the same patch attempted again, a refactor whose end state it can no longer describe, or more apologies than code edits. The lines inside are written in the second person and meant to be read in order, before the next tool call rather than after it.
A three-step reset follows. List only what has been observed: file contents, command output, error text. Then name the assumption nobody tested and run the smallest possible check of it. The rules section says to drop the abandoned theory in one clause and stop discussing it, and to give the user one sentence of acknowledgement at most.
The skill addresses the agent's state, not the problem's subject. The bug will still be there. A reset only counts when the next action differs from the previous one, so repeating the last attempt just lengthens the loop by a turn. Reaching for it on the first obstacle wastes it the other way.
What it helps with
- break out of a loop where the same fix keeps returning
- surface the assumption the agent never tested
- regain the thread halfway through a refactor
- trade repeated apologies for actual code changes
- drop an approach the agent has been defending
Example prompts for the agent
- You've patched this file three times now. Read the affirmations and restart from what the output actually says.
- We're going in circles — re-ground: what's verified, and which assumption is untested?
- Skip the apology. Name what the command output proves and test that assumption with one small step.
How to install affirmations
1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill affirmations -a claude-code2Manually: put the files in place
~/.claude/skills/affirmations/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/affirmations/
1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill affirmations -a codex2Manually: put the files in place
.agents/skills/affirmations/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill affirmations -a cursor2Manually: put the files in place
.cursor/skills/affirmations/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install 101-skills/superpowers guides/agent/affirmations2Manually: put the files in place
.github/skills/affirmations/SKILL.md1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill affirmations -a gemini-cli2Platform CLI
gemini skills install https://github.com/101-skills/superpowers --path guides/agent/affirmationsManually: put the files in place
.devin/skills/affirmations/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill affirmations -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about affirmations
How do I install affirmations in Claude Code?
Run npx skills add 101-skills/superpowers --skill affirmations -a claude-code. The skill lands at ~/.claude/skills/affirmations/SKILL.md.
What does affirmations do?
It hands the agent a short text to read while it is looping, plus a reset routine: state the verified facts, find the unchecked assumption, test it with one small action.
How can I tell the reset worked?
Watch the next action. If it matches the previous one, nothing changed; the skill only helped when the agent does something different.