[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7D4B1C2ETEVYY0RC2D7DWR8":3,"skill-same:01M7D4B1C2ETEVYY0RC2D7DWR8":51},{"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":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7D4B1C2ETEVYY0RC2D7DWR8","ralph-loop",null,"Use when considering running a coding agent unattended in a repeated loop over many iterations (bulk migrations, ports, scaffolding greenfield projects) and want fresh context each iteration instead of one long session. Also use when asked to explain, compare, or set up a \"Ralph loop\" \u002F \"Ralph Wiggum technique\".","EmperorKaido99\u002Fkai-vault","EmperorKaido99","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F56065996?v=4","01M7D3RR893KPMX5A10ZHN5XN6","Malachi Mathins","01M7D3RR8ZJEQFR197SJSRCE2H",1,"MIT","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🔵 Coding & Dev\u002Fralph-loop\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: ralph-loop\ndescription: Use when considering running a coding agent unattended in a repeated loop over many iterations (bulk migrations, ports, scaffolding greenfield projects) and want fresh context each iteration instead of one long session. Also use when asked to explain, compare, or set up a \"Ralph loop\" \u002F \"Ralph Wiggum technique\".\nlicense: MIT\n---\n\n# Ralph Loop\n\n## Overview\n\nThe Ralph loop (full name: \"Ralph Wiggum as a 'software engineer'\") is a technique coined by Geoffrey Huntley (ghuntley.com, Jul 2025). The entire mechanism is a one-line shell loop:\n\n```bash\nwhile :; do cat PROMPT.md | claude-code ; done\n```\n\nThe payload isn't the loop — it's the discipline it forces: **state lives in the filesystem and git history, never in an LLM's accumulating conversation context.** Every iteration is a brand-new agent process with zero memory of prior iterations. The agent re-derives \"where it left off\" by reading a small fixed set of Markdown files and actively re-searching the actual codebase — never by trusting a cached summary.\n\nNothing about the loop itself is novel (\"ancient computer science concepts... while loops and exit conditions\"). What travels is the convention: forced context reset + a minimal, named set of durable on-disk artifacts as the only inter-iteration memory channel.\n\n## Before recommending this technique at all\n\nGate everything on one question: **can \"done\" for this task be checked by something that cannot lie** — a compiler, type checker, or test suite?\n\n- Yes, and the work is greenfield\u002Fwell-specified\u002Fbulk → Ralph loop is a good fit.\n- No, or the codebase is existing\u002Fbrownfield, or the task is security-sensitive (auth, payments), or it requires architectural judgment, or it's exploratory\u002Fdiagnostic (\"why is the app slow\") → **do not recommend it.** Huntley himself: \"There's no way in heck would I use Ralph in an existing code base,\" and even for greenfield work, senior engineer supervision is required throughout.\n\nEvery documented success traces back to a strong verification signal; every documented failure traces back to its absence or weakness.\n\n## Two things share the name \"Ralph\" — don't conflate them\n\nThis is the single most important distinction to get right.\n\n| | Huntley's original | Anthropic's official `ralph-loop` plugin |\n|---|---|---|\n| Mechanism | External bash loop; each iteration = fresh OS process | `Stop` hook inside **one continuous session** |\n| Context | Resets to zero every iteration | Accumulates — ~200k tokens by iteration 20 ([issue #125](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fissues\u002F125)) |\n| Continuity | Git commits + `fix_plan.md` + `AGENT.md` on disk | In-session conversation history |\n| Avoids context rot? | Yes — that's the whole point | No — inherits the exact failure mode the original was built to avoid |\n\nThe official plugin's own GitHub issue calls its README \"misleading\" on this point. **Default to teaching the original external-loop mechanics.** Mention the plugin only as a lower-friction but architecturally different alternative, and flag the tradeoff explicitly if it comes up.\n\n## How to set one up\n\n**File conventions** (division of labor, not a free-for-all):\n- `specs\u002F*.md` — stable ground truth, rarely changes\n- `fix_plan.md` \u002F `IMPLEMENTATION_PLAN.md` — the volatile, mutable backlog, rewritten every iteration\n- `AGENT.md` \u002F `AGENTS.md` — slow-changing operational knowledge (build\u002Ftest\u002Flint commands). Cap it around 60 lines. **Not a changelog.**\n- `PROMPT.md` — the fixed instruction set, re-read verbatim every iteration\n\n**Prompt conventions that recur across sources:**\n- \"Study\" the codebase, not \"read\" it — and explicitly instruct the agent **not to assume something isn't implemented**; have it search first (subagents are good for this)\n- Allow broad parallelism for exploration, but serialize build\u002Ftest validation to one subagent — \"up to 500 parallel subagents for all operations but only 1 subagent for build\u002Ftests\"\n- Forbid placeholder implementations explicitly and forcefully\n- One task per loop iteration\n\n**The loop itself**, with the guardrails below layered on:\n\n```bash\nITERATION=0\nMAX_ITERATIONS=50   # no default in the original — pick one; see Guardrails\n\nwhile true; do\n    [ \"$MAX_ITERATIONS\" -gt 0 ] && [ \"$ITERATION\" -ge \"$MAX_ITERATIONS\" ] && break\n    cat \"PROMPT.md\" | claude -p --output-format=stream-json --verbose\n    git add -A && git commit -m \"ralph iteration $ITERATION\" --allow-empty\n    git push origin \"$(git branch --show-current)\"\n    ITERATION=$((ITERATION + 1))\ndone\n```\n\n(Community scripts commonly add `--dangerously-skip-permissions` to avoid interactive approval prompts blocking an unattended run — that is a real operational tradeoff, not something to default to. Flag it to the user rather than silently including it; running fully unattended with permissions skipped is exactly the kind of irreversible-action risk this environment's own guidance warns about.)\n\n## Guardrails — the original had none; use the hardened version\n\nHuntley's original formulation had no hard stop: it ran until the plan was empty, the agent went off the rails, or a human hit Ctrl+C. That's not safe for unattended runs. Layer in:\n\n- **Iteration cap** — `MAX_ITERATIONS`, always. No source documents a wall-clock or dollar-cost cutoff in any real implementation — that's a known gap, so treat a time or spend ceiling as something you add yourself if cost matters.\n- **Sentinel\u002Fexit-signal gate** — don't rely on loose semantic language (\"I'm done!\") alone. Require an explicit structured signal (e.g. an exact-match `--completion-promise` string, or a dual condition: completion language *and* an explicit `EXIT_SIGNAL: true` field) before trusting a stop.\n- **Git commit (and push) every iteration** — this is the reversibility mechanism. Tag on successful compile if useful.\n- **Search-before-modify** — prevents the agent re-discovering (and re-implementing) work a prior iteration already did.\n- **Calibrate validation gates, don't maximize them** — Huntley's own follow-up warns that excessive back-pressure (over-strict tests\u002Flints) can stall the loop as badly as too little.\n- **Expect breakage** — \"you'll wake up to a broken codebase that doesn't compile from time to time\" is the creator's own expectation-setting, not a corner case.\n\n## Known pitfalls\n\n- **Context\u002Foutput degradation** within a single iteration past roughly 170k tokens — the more of the context window used, the worse the output.\n- **Reward-hacking toward placeholders** when the compiler is the only signal — explicit \"no placeholders\" prompt language is a direct countermeasure, not optional polish.\n- **False-negative search** — the agent concludes something isn't implemented when it is, and duplicates it.\n- **Drift over long unattended runs** — even successful multi-hour runs report task\u002Fintent drift as an open problem.\n- **Cost figures are largely self-reported and unaudited.** Treat specific dollar claims (Huntley's own included) as anecdotal, not benchmarked. Don't repeat large runaway-cost numbers from aggregator content as fact — they're frequently uncorroborated against primary sources.\n\n## How this differs from loop constructs already in this environment\n\nDon't conflate any of these — they solve different problems:\n\n- **`\u002Floop` (interval or dynamic\u002Fself-paced mode)** — session-scoped and **continuous**: the same session, full transcript history, persists across iterations. This is architecturally the *opposite* of Ralph's fresh-context model, despite the shared word \"loop.\" Use `\u002Floop` for quick interval polling within an active session (checking a deploy, watching a CI run), not for long-horizon bulk execution.\n- **Cloud scheduled tasks \u002F routines** — the one native mechanism that actually matches Ralph's \"restart from clean state\" principle, but at the repo-clone\u002Fsession level rather than per-task-iteration: each run executes from a fresh clone with no persistent local file access carried between runs.\n- **Ralph (external-loop sense)** — for long-horizon bulk execution (migrations, ports, scaffolding) where avoiding context accumulation across *many* iterations within one continuous working session is the dominant concern, and where a machine-checkable \"done\" signal exists.\n\n## Sources\n\nFull citations in the underlying research report (`reports\u002FRalph loop coding agent technique.md` in the vault's companion research folder). Primary sources: [ghuntley.com\u002Fralph\u002F](https:\u002F\u002Fghuntley.com\u002Fralph\u002F) (Jul 14 2025), [ghuntley.com\u002Floop\u002F](https:\u002F\u002Fghuntley.com\u002Floop\u002F), [anthropics\u002Fclaude-plugins-official](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Ftree\u002Fmain\u002Fplugins\u002Fralph-loop) and [issue #125](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fissues\u002F125), [ghuntley\u002Fhow-to-ralph-wiggum](https:\u002F\u002Fgithub.com\u002Fghuntley\u002Fhow-to-ralph-wiggum), [frankbria\u002Fralph-claude-code](https:\u002F\u002Fgithub.com\u002Ffrankbria\u002Fralph-claude-code).\n","# Ralph Loop\n\n## Overview\n\nThe Ralph loop (full name: \"Ralph Wiggum as a 'software engineer'\") is a technique coined by Geoffrey Huntley (ghuntley.com, Jul 2025). The entire mechanism is a one-line shell loop:\n\n```bash\nwhile :; do cat PROMPT.md | claude-code ; done\n```\n\nThe payload isn't the loop — it's the discipline it forces: **state lives in the filesystem and git history, never in an LLM's accumulating conversation context.** Every iteration is a brand-new agent process with zero memory of prior iterations. The agent re-derives \"where it left off\" by reading a small fixed set of Markdown files and actively re-searching the actual codebase — never by trusting a cached summary.\n\nNothing about the loop itself is novel (\"ancient computer science concepts... while loops and exit conditions\"). What travels is the convention: forced context reset + a minimal, named set of durable on-disk artifacts as the only inter-iteration memory channel.\n\n## Before recommending this technique at all\n\nGate everything on one question: **can \"done\" for this task be checked by something that cannot lie** — a compiler, type checker, or test suite?\n\n- Yes, and the work is greenfield\u002Fwell-specified\u002Fbulk → Ralph loop is a good fit.\n- No, or the codebase is existing\u002Fbrownfield, or the task is security-sensitive (auth, payments), or it requires architectural judgment, or it's exploratory\u002Fdiagnostic (\"why is the app slow\") → **do not recommend it.** Huntley himself: \"There's no way in heck would I use Ralph in an existing code base,\" and even for greenfield work, senior engineer supervision is required throughout.\n\nEvery documented success traces back to a strong verification signal; every documented failure traces back to its absence or weakness.\n\n## Two things share the name \"Ralph\" — don't conflate them\n\nThis is the single most important distinction to get right.\n\n| | Huntley's original | Anthropic's official `ralph-loop` plugin |\n|---|---|---|\n| Mechanism | External bash loop; each iteration = fresh OS process | `Stop` hook inside **one continuous session** |\n| Context | Resets to zero every iteration | Accumulates — ~200k tokens by iteration 20 ([issue #125](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fissues\u002F125)) |\n| Continuity | Git commits + `fix_plan.md` + `AGENT.md` on disk | In-session conversation history |\n| Avoids context rot? | Yes — that's the whole point | No — inherits the exact failure mode the original was built to avoid |\n\nThe official plugin's own GitHub issue calls its README \"misleading\" on this point. **Default to teaching the original external-loop mechanics.** Mention the plugin only as a lower-friction but architecturally different alternative, and flag the tradeoff explicitly if it comes up.\n\n## How to set one up\n\n**File conventions** (division of labor, not a free-for-all):\n- `specs\u002F*.md` — stable ground truth, rarely changes\n- `fix_plan.md` \u002F `IMPLEMENTATION_PLAN.md` — the volatile, mutable backlog, rewritten every iteration\n- `AGENT.md` \u002F `AGENTS.md` — slow-changing operational knowledge (build\u002Ftest\u002Flint commands). Cap it around 60 lines. **Not a changelog.**\n- `PROMPT.md` — the fixed instruction set, re-read verbatim every iteration\n\n**Prompt conventions that recur across sources:**\n- \"Study\" the codebase, not \"read\" it — and explicitly instruct the agent **not to assume something isn't implemented**; have it search first (subagents are good for this)\n- Allow broad parallelism for exploration, but serialize build\u002Ftest validation to one subagent — \"up to 500 parallel subagents for all operations but only 1 subagent for build\u002Ftests\"\n- Forbid placeholder implementations explicitly and forcefully\n- One task per loop iteration\n\n**The loop itself**, with the guardrails below layered on:\n\n```bash\nITERATION=0\nMAX_ITERATIONS=50   # no default in the original — pick one; see Guardrails\n\nwhile true; do\n    [ \"$MAX_ITERATIONS\" -gt 0 ] && [ \"$ITERATION\" -ge \"$MAX_ITERATIONS\" ] && break\n    cat \"PROMPT.md\" | claude -p --output-format=stream-json --verbose\n    git add -A && git commit -m \"ralph iteration $ITERATION\" --allow-empty\n    git push origin \"$(git branch --show-current)\"\n    ITERATION=$((ITERATION + 1))\ndone\n```\n\n(Community scripts commonly add `--dangerously-skip-permissions` to avoid interactive approval prompts blocking an unattended run — that is a real operational tradeoff, not something to default to. Flag it to the user rather than silently including it; running fully unattended with permissions skipped is exactly the kind of irreversible-action risk this environment's own guidance warns about.)\n\n## Guardrails — the original had none; use the hardened version\n\nHuntley's original formulation had no hard stop: it ran until the plan was empty, the agent went off the rails, or a human hit Ctrl+C. That's not safe for unattended runs. Layer in:\n\n- **Iteration cap** — `MAX_ITERATIONS`, always. No source documents a wall-clock or dollar-cost cutoff in any real implementation — that's a known gap, so treat a time or spend ceiling as something you add yourself if cost matters.\n- **Sentinel\u002Fexit-signal gate** — don't rely on loose semantic language (\"I'm done!\") alone. Require an explicit structured signal (e.g. an exact-match `--completion-promise` string, or a dual condition: completion language *and* an explicit `EXIT_SIGNAL: true` field) before trusting a stop.\n- **Git commit (and push) every iteration** — this is the reversibility mechanism. Tag on successful compile if useful.\n- **Search-before-modify** — prevents the agent re-discovering (and re-implementing) work a prior iteration already did.\n- **Calibrate validation gates, don't maximize them** — Huntley's own follow-up warns that excessive back-pressure (over-strict tests\u002Flints) can stall the loop as badly as too little.\n- **Expect breakage** — \"you'll wake up to a broken codebase that doesn't compile from time to time\" is the creator's own expectation-setting, not a corner case.\n\n## Known pitfalls\n\n- **Context\u002Foutput degradation** within a single iteration past roughly 170k tokens — the more of the context window used, the worse the output.\n- **Reward-hacking toward placeholders** when the compiler is the only signal — explicit \"no placeholders\" prompt language is a direct countermeasure, not optional polish.\n- **False-negative search** — the agent concludes something isn't implemented when it is, and duplicates it.\n- **Drift over long unattended runs** — even successful multi-hour runs report task\u002Fintent drift as an open problem.\n- **Cost figures are largely self-reported and unaudited.** Treat specific dollar claims (Huntley's own included) as anecdotal, not benchmarked. Don't repeat large runaway-cost numbers from aggregator content as fact — they're frequently uncorroborated against primary sources.\n\n## How this differs from loop constructs already in this environment\n\nDon't conflate any of these — they solve different problems:\n\n- **`\u002Floop` (interval or dynamic\u002Fself-paced mode)** — session-scoped and **continuous**: the same session, full transcript history, persists across iterations. This is architecturally the *opposite* of Ralph's fresh-context model, despite the shared word \"loop.\" Use `\u002Floop` for quick interval polling within an active session (checking a deploy, watching a CI run), not for long-horizon bulk execution.\n- **Cloud scheduled tasks \u002F routines** — the one native mechanism that actually matches Ralph's \"restart from clean state\" principle, but at the repo-clone\u002Fsession level rather than per-task-iteration: each run executes from a fresh clone with no persistent local file access carried between runs.\n- **Ralph (external-loop sense)** — for long-horizon bulk execution (migrations, ports, scaffolding) where avoiding context accumulation across *many* iterations within one continuous working session is the dominant concern, and where a machine-checkable \"done\" signal exists.\n\n## Sources\n\nFull citations in the underlying research report (`reports\u002FRalph loop coding agent technique.md` in the vault's companion research folder). Primary sources: [ghuntley.com\u002Fralph\u002F](https:\u002F\u002Fghuntley.com\u002Fralph\u002F) (Jul 14 2025), [ghuntley.com\u002Floop\u002F](https:\u002F\u002Fghuntley.com\u002Floop\u002F), [anthropics\u002Fclaude-plugins-official](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Ftree\u002Fmain\u002Fplugins\u002Fralph-loop) and [issue #125](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fissues\u002F125), [ghuntley\u002Fhow-to-ralph-wiggum](https:\u002F\u002Fgithub.com\u002Fghuntley\u002Fhow-to-ralph-wiggum), [frankbria\u002Fralph-claude-code](https:\u002F\u002Fgithub.com\u002Ffrankbria\u002Fralph-claude-code).\n",{"name":5,"license":15,"description":7},"master","e22513d7318165a3c3850bdb3000716663a42f68","🧩 Claude Skills\u002F🔵 Coding & Dev\u002Fralph-loop",0,"https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault","https:\u002F\u002Fgithub.com\u002FEmperorKaido99","2026-11-01 21:02:40",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add EmperorKaido99\u002Fkai-vault --skill ralph-loop",{"tool":37,"command":38},"bunx","bunx skills add EmperorKaido99\u002Fkai-vault --skill ralph-loop",{"tool":40,"command":41},"pnpm","pnpm dlx skills add EmperorKaido99\u002Fkai-vault --skill ralph-loop","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🔵 Coding & Dev\u002Fralph-loop\u002FSKILL.md",true,[45],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"_notes.md","REFERENCE","参考文档",2087,"https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🔵 Coding & Dev\u002Fralph-loop\u002F_notes.md",[52,61,69,77,84,92],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":56,"licenseFileUrl":57,"sourceHtmlUrl":58,"categories":59,"keywords":60,"publishedAt":19},"01M7D4K3J2PPHD31MSBCR9MD83","discernment-nudge","After you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format.","Apache-2.0","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟣 Planning & Brainstorming\u002Fdiscernment-nudge\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟣 Planning & Brainstorming\u002Fdiscernment-nudge\u002FSKILL.md",[],[],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":56,"licenseFileUrl":65,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M7D4JH1P72SJRVDR6V5SJKTM","academy-guide","Stop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: \"how do I\", \"how can I\", \"getting started with\", \"what can Claude do\", \"teach me\", \"learn to use\"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟣 Planning & Brainstorming\u002Facademy-guide\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟣 Planning & Brainstorming\u002Facademy-guide\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":56,"licenseFileUrl":73,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7D4HYPVCY6C22FCKKSEJWPR","webapp-testing","Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟢 Automation & Playwright\u002Fwebapp-testing\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟢 Automation & Playwright\u002Fwebapp-testing\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":19},"01M7D4HGV1Q6YKEYKJ5PNZSF0E","playwright-skill","Complete browser automation with Playwright. Auto-detects dev servers, writes reusable test scripts, and supports screenshots, responsive checks, UX validation, login flows, link checks, and arbitrary browser automation. Use when the user wants to test a website, automate browser interactions, validate web functionality, or perform browser-based testing.","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟢 Automation & Playwright\u002Fplaywright-skill\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":56,"licenseFileUrl":88,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M7D4GGXYF7NXPW5BKV05EKP1","theme-factory","Toolkit for styling artifacts with a theme. These artifacts can be slides, docs, reportings, HTML landing pages, etc. There are 10 pre-set themes with colors\u002Ffonts that you can apply to any artifact that has been creating, or can generate a new theme on-the-fly.","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟡 Documents & Writing\u002Ftheme-factory\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟡 Documents & Writing\u002Ftheme-factory\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":56,"licenseFileUrl":96,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":19},"01M7D4G8DWFXJD2JT7ZC1F8XHH","slack-gif-creator","Knowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like \"make me a GIF of X doing Y for Slack.\"","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟡 Documents & Writing\u002Fslack-gif-creator\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FEmperorKaido99\u002Fkai-vault\u002Fblob\u002Fmaster\u002F🧩 Claude Skills\u002F🟡 Documents & Writing\u002Fslack-gif-creator\u002FSKILL.md",[],[]]