Swing Skills
Open-source Claude Code skills — 6 cognitive firewalls block AI hallucination, bias & sloppy reasoning. npx skills add
Install
npx skills add thestack-ai/swing-skills --allclaude mcp add swing-skills -- npx skills add thestack-ai/swing-skills --all{
"mcpServers": {
"swing-skills": {
"command": "npx",
"args": [
"skills",
"add",
"thestack-ai/swing-skills",
"--all"
]
}
}
}Install command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →Swing — AI Agent Guardrails for Claude Code
English · 한국어 (Korean)
Your AI is not swinging at all.
It picks the safe default, agrees when it should push back, cites no sources, hides its reasoning, and assumes everything will work. Six cognitive failures. Six firewalls. Make it swing.
npx skills add thestack-ai/swing-skills --all
Built for Claude Code | Compatible with AI agents via Agent Skills open standard Created by @thestack_ai
See the Difference
Before: Default AI Research
You: Is SQLite viable for an app with 1000 concurrent users?
AI: No, SQLite is not suitable for high-concurrency applications. Use PostgreSQL or MySQL instead for production workloads.
After: With swing-research
You: /swing-research Is SQLite viable for an app with 1000 concurrent users?
AI: Yes, with caveats. 1000 concurrent users ≠ 1000 concurrent writes. Typical web apps generate ~30 concurrent write transactions at this scale. SQLite handles ~120 write transactions/sec in WAL mode (verified: sqlite.org benchmarks + Expensify production case study serving 10M+ users on SQLite). The real constraint is write contention above ~100 concurrent writers, not user count.
Sources: 🏛️ sqlite.org WAL documentation, 🛡️ Expensify engineering blog (2024), ⚠️ HN discussion (corroborated)
The baseline gave a wrong answer confidently. The skill forced source verification and reached the opposite — correct — conclusion.
Before: Default AI Code Review
You: Review this JWT auth module
AI: The code looks well-structured. Consider adding more comments and using TypeScript for better type safety. Variable naming could be improved in some places.
After: With swing-review
You: /swing-review Review this JWT auth module
AI: Steel-man: JWT-based stateless auth was chosen for horizontal scaling. The 15-min access + 7-day refresh split balances security and UX.
🔴 Critical: No refresh token rotation. A stolen token grants persistent 7-day access with no detection. Fix: implement token family chain rotation.
🟠 Major: Rate limiting uses in-memory Map(). Resets on restart, bypassed in multi-instance deploys. Fix: move to Redis with TTL.
Verdict: PASS WITH CONDITIONS — must fix token rotation before deploy.
The baseline found nothing. The skill found a Critical security vulnerability.
Six Failures, Six Firewalls
Similar plugs
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local…
Why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
Production-grade engineering skills for AI coding agents.
The 1st Collaborative AI Design Agent Workspace. Local-first and open source. Create prototypes, websites, apps, dashboards, presentations, images, and videos with your team…