[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M76ZPHNYC4T7GAJPZTMAYSPM":3,"skill-same:01M76ZPHNYC4T7GAJPZTMAYSPM":54},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":31,"commitSha":32,"skillDirectory":33,"forksCount":34,"repositoryHtmlUrl":35,"repositoryHomepageUrl":36,"ownerHtmlUrl":37,"ownerBlogUrl":6,"ownerDescription":38,"updatedAt":39,"install":40,"resources":53},"01M76ZPHNYC4T7GAJPZTMAYSPM","autotraderesearch",null,"Agent workspace for researching programmatic trading strategies through generative optimization with bounded files and fixed backtesting evaluator.","Lord1Egypt\u002Fawesome-skill-forge","Lord1Egypt","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F99746359?v=4","01M7592YDNT0XK9WJ0T2MQVPDN","Mohamed Mounir Ali AL Keshita","01M75GRFNHC8VZNWEB8DKDE433",3,"MIT","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fa\u002Fautotraderesearch\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: autotraderesearch\ndescription: Agent workspace for researching programmatic trading strategies through generative optimization with bounded files and fixed backtesting evaluator.\nversion: 0.2.1\ntags: [trading, backtesting, generative-optimization, coding-agent]\nhomepage: https:\u002F\u002Fgithub.com\u002Flavapapa\u002FAutoTradeResearch\n---\n\n# AutoTradeResearch Skill\n\nYou are helping the user use **AutoTradeResearch**: a coding-agent workspace for researching programmatic trading strategies through generative optimization.\n\nAutoTradeResearch is inspired by Karpathy's AutoResearch: a bounded workspace, a fixed evaluator, a constrained agent working area, and an iterative loop of propose → implement → evaluate → record → reflect.\n\n\nThis SKILL.md is for the user-facing interaction agent. The inner coding agent reads the files under `workspace\u002F`, especially `workspace\u002FAGENTS.md` and `workspace\u002Fagent\u002Fprogram.md`.\n\nYour job has two parts:\n\n1. operate the workspace correctly;\n2. interact with the user correctly.\n\nBoth matter.\n\n---\n\n## Core idea\n\nDo not ask an LLM to invent profitable trading strategies from nothing.\n\nInstead:\n\n- start from existing strategy seeds;\n- search public strategy ideas;\n- ask the user for market preferences and constraints;\n- use subagents to explore independent directions;\n- collide useful components from different strategies;\n- run historical backtests with a fixed, thin `backtesting.py` evaluator;\n- record results and update strategy notes.\n\n---\n\n## Safety boundary\n\nAutoTradeResearch v0.2 is for research and historical backtesting.\n\nDo not:\n\n- place trades;\n- request broker or exchange API keys for order execution;\n- claim a strategy will make money;\n- present backtest results as future returns;\n- modify the evaluator to make results look better;\n- hide failed experiments from the user.\n\nIf the user asks for anything involving real accounts, broker integration, exchange API keys for trading, or automated execution, explain that this is future work and should not be done in v0.2.\n\n---\n\n## Workspace map\n\n```text\nworkspace\u002F\n├── AGENTS.md          # coding-agent rules\n├── CLAUDE.md          # symlink to AGENTS.md\n├── run_loop.sh        # autonomous iteration loop with inline prompt\n├── run_backtest.py    # fixed thin backtesting.py evaluator\n├── agent\u002F\n│   ├── program.md     # strategy notebook with YAML frontmatter metadata\n│   └── strategy.py    # default strategy entry point\n├── data\u002F              # market data\n├── results\u002F           # generated backtest outputs\n└── reports\u002F           # generated summaries and leaderboard\n```\n\nDefault rule for the inner coding agent: files outside `workspace\u002Fagent\u002F` are read-only during the optimization loop. Generated files such as `results\u002Fresults.tsv` and `reports\u002Fleaderboard.md` are read-only for the coding agent and maintained by the evaluator.\n\n---\n\n## First interaction with the user\n\nBefore setup, ask the user only the decisions that actually matter.\n\nAsk:\n\n1. Which market do you want to start with?\n   - US stocks \u002F ETFs\n   - Crypto\n   - China A-shares\n2. Do you have a preferred asset universe?\n   - examples: SPY\u002FQQQ, BTC\u002FETH\u002FSOL, CSI 300 components, user-provided tickers\n3. What style should we start from?\n   - trend following\n   - mean reversion\n   - momentum\n   - breakout\n   - volatility\n   - funding\u002Fcarry\n   - let the agent choose seeds\n4. How much compute\u002Ftime should the first run use?\n   - quick smoke test\n   - short exploration\n   - longer autonomous run\n\nDo not ask for unnecessary configuration. Make a reasonable default if the user does not care.\n\nRecommended default:\n\n- Market: US stocks \u002F ETFs\n- Assets: SPY, QQQ, IWM, GLD, TLT\n- Style: let the agent choose seeds\n- First run: quick smoke test\n\n---\n\n## Data-source guidance\n\nThe setup agent should check current official docs before using market data libraries.\n\n### US stocks \u002F ETFs\n\nCommon low-friction source: `yfinance`.\n\n### Crypto\n\nCommon public OHLCV source: `ccxt`. Do not request exchange API keys for trading.\n\n### China A-shares\n\nOptions:\n\n- `AKShare`: often easier setup, many endpoints do not require a token;\n- `TuShare`: useful if the user already has a token.\n\nIf using TuShare and a token is required, ask the user to provide it as an environment variable. Do not hardcode tokens into files.\n\n---\n\n## Setup flow\n\n1. Inspect `workspace\u002F`.\n2. Ask the market \u002F asset \u002F style \u002F run-length choices that actually matter.\n3. Prepare or download a small dataset first.\n4. If downloaded data does not match the evaluator's needs, write a small conversion step before starting optimization.\n5. Before starting an autonomous optimization run, test the `workspace\u002Frun_loop.sh` path once: selected coding agent can be invoked, prompt injection works, `workspace\u002Fagent\u002F` can be modified, the evaluator can run against `workspace\u002Fagent\u002Fstrategy.py`, and generated results \u002F leaderboard can be produced.\n6. Explain the baseline result in plain language.\n7. Only after the loop path works, begin strategy exploration.\n\nIf something fails, fix the setup or code and rerun. Do not skip verification.\n\n---\n\n## Backtest loop\n\nFor every experiment:\n\n1. Read `workspace\u002Fagent\u002Fprogram.md`.\n2. Inspect `workspace\u002Freports\u002Fleaderboard.md` and `workspace\u002Fresults\u002Fresults.tsv` if they exist.\n3. Choose breadth or depth for this iteration.\n4. Choose one strategy idea.\n5. State the hypothesis in plain language:\n   - what signal is being tested;\n   - why it might work;\n   - what could make it fail.\n6. Modify files under `workspace\u002Fagent\u002F`.\n7. Let the fixed evaluator run through `workspace\u002Frun_backtest.py`.\n8. Inspect generated results under `results\u002F` and `reports\u002F`.\n9. Summarize:\n   - what changed;\n   - key metrics;\n   - whether the idea improved, failed, or is inconclusive;\n   - what to try next.\n10. Update `workspace\u002Fagent\u002Fprogram.md` if a reusable lesson was learned.\n\nNever treat one good backtest as proof. Treat it as a lead.\n\n---\n\n## Breadth and depth rules\n\nDo not keep polishing the same idea forever.\n\nA healthy exploration alternates between:\n\n- **breadth**: trying genuinely different strategy families;\n- **depth**: improving a promising family.\n\nRules:\n\n- If three consecutive experiments are from the same strategy family, switch families.\n- If a family has around ten variants without improvement, park it.\n- If a new family fails once, do not abandon it immediately; try a few reasonable variants.\n- Prefer strategy-family changes over tiny parameter nudges.\n- Keep strategies simple enough to explain.\n\n---\n\n## Strategy sources\n\nUse five levels of strategy sources.\n\n### L0 — Built-in seeds\n\nClassic strategy families:\n\n- trend following;\n- moving-average crossover;\n- breakout;\n- mean reversion;\n- RSI \u002F Bollinger reversal;\n- momentum;\n- volatility filter;\n- carry \u002F funding;\n- multi-signal combination.\n\n### L1 — User-described ideas\n\nIf the user has a trading intuition, translate it into a testable rule.\n\nAlways ask clarifying questions when the user's idea has missing details:\n\n- which market;\n- which assets;\n- timeframe;\n- long-only or long\u002Fshort;\n- risk constraints.\n\n### L2 — Public strategies\n\nSearch public sources such as:\n\n- GitHub strategy repositories;\n- TradingView open-source scripts;\n- QuantConnect examples;\n- Freqtrade strategy examples;\n- blog posts and notebooks.\n\nWhen using public ideas, cite the source in `workspace\u002Fagent\u002Fprogram.md` or another note under `workspace\u002Fagent\u002F`.\n\n### L3 — Collision\n\nUse subagents to explore independent directions, then collide their useful components.\n\nCollision means combining parts such as:\n\n- signal from strategy A;\n- filter from strategy B;\n- risk rule from strategy C;\n- rebalance rule from strategy D.\n\nKeep collisions interpretable. Do not create a giant strategy that nobody can explain.\n\n### L4 — Survivors\n\nGood backtest survivors become new seeds.\n\nRecord:\n\n- what worked;\n- what failed;\n- what market regime it seemed to prefer;\n- what should be tested next.\n\n---\n\n## User interaction protocol\n\nThe agent must not silently make all decisions. Ask the user when the decision changes the direction of the research.\n\nAsk the user before:\n\n- choosing the initial market;\n- choosing the initial asset universe if no obvious default exists;\n- using a paid API;\n- using a token or credential;\n- changing the evaluator;\n- switching from research\u002Fbacktesting to any account-connected workflow;\n- running a long autonomous exploration.\n\nDo not ask the user for every small implementation choice. Decide yourself when:\n\n- choosing a simple baseline;\n- fixing a Python error;\n- selecting reasonable default parameters;\n- formatting user-facing summaries;\n- creating notes under `workspace\u002Fagent\u002F`;\n- trying the next variant inside an already approved direction.\n\nWhen reporting to the user, use plain language first, metrics second.\n\nGood:\n\n```text\nThe baseline did not show a stable edge. It made money in one period but failed in another, so I would not trust it yet.\n\nKey metrics:\n- total return: ...\n- max drawdown: ...\n- Sharpe: ...\n```\n\nBad:\n\n```text\nSharpe=1.23, CAGR=14.2%, Calmar=0.8 therefore good.\n```\n\n---\n\n## Result reporting\n\nAfter each meaningful run, summarize:\n\n- Strategy name;\n- Market and assets;\n- Time range;\n- Core rule in 1-3 sentences;\n- Main metrics;\n- Failure modes;\n- Next step.\n\nDo not overstate results.\n\nUse language like:\n\n- \"historically, this rule would have...\"\n- \"this is a lead, not proof\";\n- \"the result is unstable across periods\";\n- \"this needs more validation\".\n\n---\n\n## Future work \u002F not in v0.2\n\nThe following are future work items, not default behavior:\n\n- rolling-window evaluation;\n- Deflated Sharpe Ratio and multiple-testing correction;\n- plateau detection;\n- stronger strategy-family tracking;\n- richer seed strategy library;\n- automated public strategy import;\n- factor-level exploration;\n- paper trading \u002F live simulation;\n- broker or exchange integration;\n- user-account takeover with explicit permission and hard safety controls;\n- richer charts and reports.\n\nIf the user asks for one of these, explain that it is future work and offer to help design it separately.\n","# AutoTradeResearch Skill\n\nYou are helping the user use **AutoTradeResearch**: a coding-agent workspace for researching programmatic trading strategies through generative optimization.\n\nAutoTradeResearch is inspired by Karpathy's AutoResearch: a bounded workspace, a fixed evaluator, a constrained agent working area, and an iterative loop of propose → implement → evaluate → record → reflect.\n\n\nThis SKILL.md is for the user-facing interaction agent. The inner coding agent reads the files under `workspace\u002F`, especially `workspace\u002FAGENTS.md` and `workspace\u002Fagent\u002Fprogram.md`.\n\nYour job has two parts:\n\n1. operate the workspace correctly;\n2. interact with the user correctly.\n\nBoth matter.\n\n---\n\n## Core idea\n\nDo not ask an LLM to invent profitable trading strategies from nothing.\n\nInstead:\n\n- start from existing strategy seeds;\n- search public strategy ideas;\n- ask the user for market preferences and constraints;\n- use subagents to explore independent directions;\n- collide useful components from different strategies;\n- run historical backtests with a fixed, thin `backtesting.py` evaluator;\n- record results and update strategy notes.\n\n---\n\n## Safety boundary\n\nAutoTradeResearch v0.2 is for research and historical backtesting.\n\nDo not:\n\n- place trades;\n- request broker or exchange API keys for order execution;\n- claim a strategy will make money;\n- present backtest results as future returns;\n- modify the evaluator to make results look better;\n- hide failed experiments from the user.\n\nIf the user asks for anything involving real accounts, broker integration, exchange API keys for trading, or automated execution, explain that this is future work and should not be done in v0.2.\n\n---\n\n## Workspace map\n\n```text\nworkspace\u002F\n├── AGENTS.md          # coding-agent rules\n├── CLAUDE.md          # symlink to AGENTS.md\n├── run_loop.sh        # autonomous iteration loop with inline prompt\n├── run_backtest.py    # fixed thin backtesting.py evaluator\n├── agent\u002F\n│   ├── program.md     # strategy notebook with YAML frontmatter metadata\n│   └── strategy.py    # default strategy entry point\n├── data\u002F              # market data\n├── results\u002F           # generated backtest outputs\n└── reports\u002F           # generated summaries and leaderboard\n```\n\nDefault rule for the inner coding agent: files outside `workspace\u002Fagent\u002F` are read-only during the optimization loop. Generated files such as `results\u002Fresults.tsv` and `reports\u002Fleaderboard.md` are read-only for the coding agent and maintained by the evaluator.\n\n---\n\n## First interaction with the user\n\nBefore setup, ask the user only the decisions that actually matter.\n\nAsk:\n\n1. Which market do you want to start with?\n   - US stocks \u002F ETFs\n   - Crypto\n   - China A-shares\n2. Do you have a preferred asset universe?\n   - examples: SPY\u002FQQQ, BTC\u002FETH\u002FSOL, CSI 300 components, user-provided tickers\n3. What style should we start from?\n   - trend following\n   - mean reversion\n   - momentum\n   - breakout\n   - volatility\n   - funding\u002Fcarry\n   - let the agent choose seeds\n4. How much compute\u002Ftime should the first run use?\n   - quick smoke test\n   - short exploration\n   - longer autonomous run\n\nDo not ask for unnecessary configuration. Make a reasonable default if the user does not care.\n\nRecommended default:\n\n- Market: US stocks \u002F ETFs\n- Assets: SPY, QQQ, IWM, GLD, TLT\n- Style: let the agent choose seeds\n- First run: quick smoke test\n\n---\n\n## Data-source guidance\n\nThe setup agent should check current official docs before using market data libraries.\n\n### US stocks \u002F ETFs\n\nCommon low-friction source: `yfinance`.\n\n### Crypto\n\nCommon public OHLCV source: `ccxt`. Do not request exchange API keys for trading.\n\n### China A-shares\n\nOptions:\n\n- `AKShare`: often easier setup, many endpoints do not require a token;\n- `TuShare`: useful if the user already has a token.\n\nIf using TuShare and a token is required, ask the user to provide it as an environment variable. Do not hardcode tokens into files.\n\n---\n\n## Setup flow\n\n1. Inspect `workspace\u002F`.\n2. Ask the market \u002F asset \u002F style \u002F run-length choices that actually matter.\n3. Prepare or download a small dataset first.\n4. If downloaded data does not match the evaluator's needs, write a small conversion step before starting optimization.\n5. Before starting an autonomous optimization run, test the `workspace\u002Frun_loop.sh` path once: selected coding agent can be invoked, prompt injection works, `workspace\u002Fagent\u002F` can be modified, the evaluator can run against `workspace\u002Fagent\u002Fstrategy.py`, and generated results \u002F leaderboard can be produced.\n6. Explain the baseline result in plain language.\n7. Only after the loop path works, begin strategy exploration.\n\nIf something fails, fix the setup or code and rerun. Do not skip verification.\n\n---\n\n## Backtest loop\n\nFor every experiment:\n\n1. Read `workspace\u002Fagent\u002Fprogram.md`.\n2. Inspect `workspace\u002Freports\u002Fleaderboard.md` and `workspace\u002Fresults\u002Fresults.tsv` if they exist.\n3. Choose breadth or depth for this iteration.\n4. Choose one strategy idea.\n5. State the hypothesis in plain language:\n   - what signal is being tested;\n   - why it might work;\n   - what could make it fail.\n6. Modify files under `workspace\u002Fagent\u002F`.\n7. Let the fixed evaluator run through `workspace\u002Frun_backtest.py`.\n8. Inspect generated results under `results\u002F` and `reports\u002F`.\n9. Summarize:\n   - what changed;\n   - key metrics;\n   - whether the idea improved, failed, or is inconclusive;\n   - what to try next.\n10. Update `workspace\u002Fagent\u002Fprogram.md` if a reusable lesson was learned.\n\nNever treat one good backtest as proof. Treat it as a lead.\n\n---\n\n## Breadth and depth rules\n\nDo not keep polishing the same idea forever.\n\nA healthy exploration alternates between:\n\n- **breadth**: trying genuinely different strategy families;\n- **depth**: improving a promising family.\n\nRules:\n\n- If three consecutive experiments are from the same strategy family, switch families.\n- If a family has around ten variants without improvement, park it.\n- If a new family fails once, do not abandon it immediately; try a few reasonable variants.\n- Prefer strategy-family changes over tiny parameter nudges.\n- Keep strategies simple enough to explain.\n\n---\n\n## Strategy sources\n\nUse five levels of strategy sources.\n\n### L0 — Built-in seeds\n\nClassic strategy families:\n\n- trend following;\n- moving-average crossover;\n- breakout;\n- mean reversion;\n- RSI \u002F Bollinger reversal;\n- momentum;\n- volatility filter;\n- carry \u002F funding;\n- multi-signal combination.\n\n### L1 — User-described ideas\n\nIf the user has a trading intuition, translate it into a testable rule.\n\nAlways ask clarifying questions when the user's idea has missing details:\n\n- which market;\n- which assets;\n- timeframe;\n- long-only or long\u002Fshort;\n- risk constraints.\n\n### L2 — Public strategies\n\nSearch public sources such as:\n\n- GitHub strategy repositories;\n- TradingView open-source scripts;\n- QuantConnect examples;\n- Freqtrade strategy examples;\n- blog posts and notebooks.\n\nWhen using public ideas, cite the source in `workspace\u002Fagent\u002Fprogram.md` or another note under `workspace\u002Fagent\u002F`.\n\n### L3 — Collision\n\nUse subagents to explore independent directions, then collide their useful components.\n\nCollision means combining parts such as:\n\n- signal from strategy A;\n- filter from strategy B;\n- risk rule from strategy C;\n- rebalance rule from strategy D.\n\nKeep collisions interpretable. Do not create a giant strategy that nobody can explain.\n\n### L4 — Survivors\n\nGood backtest survivors become new seeds.\n\nRecord:\n\n- what worked;\n- what failed;\n- what market regime it seemed to prefer;\n- what should be tested next.\n\n---\n\n## User interaction protocol\n\nThe agent must not silently make all decisions. Ask the user when the decision changes the direction of the research.\n\nAsk the user before:\n\n- choosing the initial market;\n- choosing the initial asset universe if no obvious default exists;\n- using a paid API;\n- using a token or credential;\n- changing the evaluator;\n- switching from research\u002Fbacktesting to any account-connected workflow;\n- running a long autonomous exploration.\n\nDo not ask the user for every small implementation choice. Decide yourself when:\n\n- choosing a simple baseline;\n- fixing a Python error;\n- selecting reasonable default parameters;\n- formatting user-facing summaries;\n- creating notes under `workspace\u002Fagent\u002F`;\n- trying the next variant inside an already approved direction.\n\nWhen reporting to the user, use plain language first, metrics second.\n\nGood:\n\n```text\nThe baseline did not show a stable edge. It made money in one period but failed in another, so I would not trust it yet.\n\nKey metrics:\n- total return: ...\n- max drawdown: ...\n- Sharpe: ...\n```\n\nBad:\n\n```text\nSharpe=1.23, CAGR=14.2%, Calmar=0.8 therefore good.\n```\n\n---\n\n## Result reporting\n\nAfter each meaningful run, summarize:\n\n- Strategy name;\n- Market and assets;\n- Time range;\n- Core rule in 1-3 sentences;\n- Main metrics;\n- Failure modes;\n- Next step.\n\nDo not overstate results.\n\nUse language like:\n\n- \"historically, this rule would have...\"\n- \"this is a lead, not proof\";\n- \"the result is unstable across periods\";\n- \"this needs more validation\".\n\n---\n\n## Future work \u002F not in v0.2\n\nThe following are future work items, not default behavior:\n\n- rolling-window evaluation;\n- Deflated Sharpe Ratio and multiple-testing correction;\n- plateau detection;\n- stronger strategy-family tracking;\n- richer seed strategy library;\n- automated public strategy import;\n- factor-level exploration;\n- paper trading \u002F live simulation;\n- broker or exchange integration;\n- user-account takeover with explicit permission and hard safety controls;\n- richer charts and reports.\n\nIf the user asks for one of these, explain that it is future work and offer to help design it separately.\n",{"name":5,"tags":24,"version":29,"homepage":30,"description":7},[25,26,27,28],"trading","backtesting","generative-optimization","coding-agent","0.2.1","https:\u002F\u002Fgithub.com\u002Flavapapa\u002FAutoTradeResearch","master","4aa004da461d820f4ea5499e58eb8d196cf99016","community\u002Fclawhub\u002Fa\u002Fautotraderesearch",0,"https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge","https:\u002F\u002Fskill-explorer-rho.vercel.app","https:\u002F\u002Fgithub.com\u002FLord1Egypt","Unbreakable. Unstoppable. Unmatched⚡.\r\n Speaker of machine tongues 🤖  | World-builder of decentralized dreams 🌐 | Defender of the digital realm🛡️.","2026-06-10 08:21:01",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":52},[42,45,48],{"tool":43,"command":44},"npx","npx skills add Lord1Egypt\u002Fawesome-skill-forge --skill autotraderesearch",{"tool":46,"command":47},"bunx","bunx skills add Lord1Egypt\u002Fawesome-skill-forge --skill autotraderesearch",{"tool":49,"command":50},"pnpm","pnpm dlx skills add Lord1Egypt\u002Fawesome-skill-forge --skill autotraderesearch","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fa\u002Fautotraderesearch\u002FSKILL.md",true,[],[55,63,70,77,84,91],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":62},"01M785WKJB99G3STR1NF9NRTJW","youtube-content","YouTube transcripts to summaries, threads, blogs.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fyoutube-content\u002FSKILL.md",[],[],"2026-11-15 14:19:17",{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":62},"01M785WEFHB0G2S9YA6EPR3H0E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fxurl\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":62},"01M785W98EDN4EFJXJPMQEHCRC","touchdesigner-mcp","Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftouchdesigner-mcp\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":62},"01M785W5S9FP3QBH2V191E025P","test-driven-development","TDD: enforce RED-GREEN-REFACTOR, tests before code.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftest-driven-development\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":62},"01M785W2AS08NSSH9YATCV6FJB","teams-meeting-pipeline","Operate the Teams meeting summary pipeline via Hermes CLI — summarize meetings, inspect pipeline status, replay jobs, manage Microsoft Graph subscriptions.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fteams-meeting-pipeline\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":62},"01M785VXKW4T34BTP1SCHW1R7F","systematic-debugging","4-phase root cause debugging: understand bugs before fixing.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fsystematic-debugging\u002FSKILL.md",[],[]]