Anti-Sycophancy Check
Makes the agent test your claims against evidence instead of simply agreeing with you
- Repository
- sickn33/agentic-awesome-skills
- Stars
- 49,023
- Licence
- MIT
- Files
- 2 files
What it does
When this skill is active, the AI assistant restates your claim in plain terms, assesses it on the evidence without deferring to your framing or authority, and gives its conclusion before its reasoning. If you push back, it checks whether you brought new evidence or just repeated your opinion, and changes its position only for the former. It changes how the assistant responds, not what information it can access.
Great for
- Developers who want a second opinion on technical decisions rather than validation
- Users who notice AI assistants being overly agreeable and want more candid answers
- People reviewing code, architecture or plans who want claims challenged
Skip it if
- Situations where you simply want help executing a task, since the skill can make responses more confrontational than needed
- Cases where the claim is already well supported; the skill itself warns against reflexive contrarianism
- Factual questions needing current or external information, because the skill does not add data sources and the assistant still needs evidence from its tools
Try asking
- “Load anti-sycophancy and tell me whether my plan to store sessions in localStorage is secure”
- “Use anti-sycophancy to assess whether my claim that this query is the bottleneck holds up”
- “With anti-sycophancy active, review my argument that we should rewrite the service in Rust”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 2 original files, the MIT licence and an attribution note.
anti-sycophancy.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip anti-sycophancy.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip anti-sycophancy.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Provide concrete evidence such as logs, benchmarks or docs when you disagree, since the skill updates its position only on new evidence
- Ask for the evidence behind a conclusion if the reasoning seems thin
- Turn the skill off for brainstorming tasks where you want open-ended, supportive input
What's inside
- README.md
- SKILL.md
Redistributed unmodified from sickn33/agentic-awesome-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.