runesleo

Project knowledge accumulation system - learn from practice, avoid repeating mistakes

File-based planning for complex tasks. Creates task_plan.md, findings.md, and progress.md. Use for multi-step tasks requiring >5 tool calls.

Session wrap-up - update handoff + commit + auto-record experience

Systematic debugging - four-phase process, find root cause before fixing

Verification before completion - must run verification commands before claiming done. Evidence before claims.

Precise PnL calculator for any Polymarket address — reconstructs profit/loss from Data API cashflow (BUY/SELL/REDEEM/MERGE/SPLIT/REBATE) plus unrealized position value. Matches official /profit within ~0.2% MAPE on leaderboard validation. Bring-your-own Python (httpx).

Brier Score calculator for Polymarket addresses — measures prediction accuracy independent of PnL. Separates skilled predictors from market makers and arbitrageurs.

Polymarket address profiler — input any 0x address, get a complete trading profile with PnL, win rate, positions, category breakdown, and top trades. All data from public APIs, no local database needed.

扫描 ~/.claude/skills/ 输出热度/死 skill/冲突/过期 pattern 报告。周日运营日跑。触发：/skill-audit、"技能树体检"、"skill 健康度"、"哪些 skill 没用"。

去 AI 味 + 让文字读起来像你自己说话。两个模式——build-my-voice（从你自己的说话样本生成一份"你的声音层"清单）和 deslop（按三轮把一篇稿子的 AI 味刮掉：去统计指纹→改字可读→对齐你的声音）。Use when 用户说 去AI味 / humanize / 让AI写的像我 / 建我的声音层 / deslop / make it sound like me，或要把 AI 初稿改成不像 AI、像本人口吻。NOT when 只是普通润色/翻译/扩写，或想绕过 AI 检测器（本 skill 不做 detector-evasion，只做"像真人、像你"）。

Remove signs of AI-generated writing from text. Detects and fixes 24 common AI patterns including inflated symbolism, promotional language, superficial analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and sycophantic tone. Based on Wikipedia's "Signs of AI writing" guide maintained by WikiProject AI Cleanup.

Four-phase debugging framework that ensures root cause investigation before attempting fixes. Never jump to solutions.

4-stage content pipeline orchestrator: Research -> Ideate -> Write -> Queue. Give it a topic, it researches existing discussions, generates hook angles, writes a draft, and queues it for review. Inspired by @shannholmberg's 4-Agent content system (Research -> Ideate -> Write -> Orchestrate). Designed for creators who build in public and want systematic content production.