[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C8TG0KA4FXSSC8R75A50TZ":3,"skill-same:01M7C8TG0KA4FXSSC8R75A50TZ":52},{"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":16,"sourceHtmlUrl":17,"categories":18,"keywords":19,"publishedAt":20,"linkOnly":21,"markdownContent":22,"markdownBody":23,"frontMatter":24,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M7C8TG0KA4FXSSC8R75A50TZ","claude",null,"Use Claude Code as an independent `claude -p` subagent when the user explicitly asks for Claude, wants a second-agent opinion from Claude, or asks to delegate a well-scoped task to Claude. Supports selecting `--model` and thinking\u002Feffort level with defaults of the current `opus` alias and `high`.","transitive-bullshit\u002Fskills","transitive-bullshit","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F552829?v=4","01M7C8KS4F4572TZFKR12N2A3W","Travis Fischer","01M7C8KS56YA2KDQHDNXEGS23P",4,"MIT","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fclaude\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fclaude\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: claude\ndescription: Use Claude Code as an independent `claude -p` subagent when the user explicitly asks for Claude, wants a second-agent opinion from Claude, or asks to delegate a well-scoped task to Claude. Supports selecting `--model` and thinking\u002Feffort level with defaults of the current `opus` alias and `high`.\n---\n\n# Claude\n\nClaude Code is an independent agent on PATH (`claude`), sharing this working\ntree and already authenticated. Use it as a second opinion or delegated worker,\nnot as ground truth: verify its claims, own any changes, and keep its task\ncontract narrow.\n\n## Defaults\n\nUse `claude -p --model opus --effort high` unless the user or task calls for a\ndifferent model or thinking level. Keep `opus` as an alias so Claude Code\nresolves the current Opus release; do not pin a dated model ID in this skill.\n\n- **Model:** prefer aliases such as `opus`, `sonnet`, or `fable`, which follow\n  the current Claude releases. Accept a full model name only when the user or\n  task requires that specific model, and verify it is current first. Default\n  `opus`.\n- **Thinking level:** pass Claude Code's `--effort` flag. Valid levels are\n  `low`, `medium`, `high`, `xhigh`, and `max`. Default `high`.\n- **Cost guard:** for exploratory asks, add `--max-budget-usd \u003Camount>` when\n  the user gives a budget or the task is likely to sprawl.\n\nExample:\n\n```sh\nclaude -p --model opus --effort high \"\u003Cprompt>\"\n```\n\n## Prompting Claude\n\nPrompt Claude like an operator: compact, block-structured, and explicit about\nthe artifact you need. A sharper contract beats higher effort.\n\n- **One task per run.** Split review, implementation, and docs into separate\n  prompts.\n- **Name installed repo skills when relevant.** Claude can read repo files;\n  say \"follow codebase-design for module boundaries\" or \"follow write-skills\n  for the skill\" instead of pasting those skills into the prompt.\n- **Blocks, added only where useful:**\n  - `\u003Ctask>` — the concrete job, context, and expected end state.\n  - `\u003Coutput_contract>` — exact response shape, highest-value first.\n  - `\u003Cdefault_follow_through>` — take low-risk interpretations and continue;\n    stop only when a missing detail changes correctness, safety, or an\n    irreversible action.\n  - `\u003Cverification_loop>` — inspect the result against requirements and revise\n    before finalizing.\n  - `\u003Cgrounding>` — ground every claim in code or tool output; label\n    inferences.\n  - `\u003Caction_safety>` — keep the diff tightly scoped; no drive-by refactors.\n- **Anti-patterns:** vague prompts, mixed jobs, no output contract, asking it\n  to \"think harder\" instead of tightening the task, and accepting its answer\n  without checking the evidence.\n\n## Consultation\n\nUse Claude for consultation when the user asks for Claude's view or when a\nsecond-agent read from Claude is already authorized. Keep it read-oriented unless\nthe user explicitly asks it to edit.\n\n1. Ask for evidence, not vibes: require file paths, observed behavior, command\n   output, or labelled inferences.\n2. Prefer a read-only tool surface for reviews:\n   `--tools \"Read,Grep,Glob\" --permission-mode dontAsk`.\n3. **Never grant fewer tools than your output contract requires.** Read your\n   own prompt and ask which tool produces each claim you demanded. A judge that\n   cannot run a command will still answer a question posed in numbers — with\n   numbers, self-consistent and measured from nothing, and with the artifacts\n   it says it wrote. Trimming the tool surface does not make the answer\n   cautious; it makes the fabrication harder to spot. Either grant the tool or\n   stop asking for the claim. Add command execution only within the authorized\n   scope and the host’s normal approval controls; otherwise narrow the contract.\n4. Require the receipt with the number: any quantity in the response must\n   arrive with the command that produced it and that command's output. Findings\n   that fail this are struck, not discounted.\n5. Triage every finding against the code before acting. Report what Claude\n   flagged, what you accepted, and what you dismissed.\n\n## Implementation — explicit ask only\n\nDelegate implementation to Claude only when the user names Claude for the task.\nKeep delegation within the authorized task. In-scope follow-ups retain that\nauthorization; ask again only when the proposed delegation expands its scope.\n\n1. Slice the task sharply: goal, constraints, allowed scope, and verification.\n   An underspecified task stays with you until a fresh agent could not misread\n   it.\n2. Start from a clean tree or record the baseline commit so Claude's diff is\n   separable from yours. Tell Claude to keep scratch files (probes, dumps,\n   notes) out of the tree — its diff must contain only the deliverable.\n3. Choose permissions by what the task must do:\n   - Read\u002Freview: `--tools \"Read,Grep,Glob\" --permission-mode dontAsk`.\n   - Narrow edits: add edit tools and keep the prompt's allowed paths explicit.\n   - Broader work: use a dedicated worktree and the host’s normal approval\n     controls. If a required action is blocked, request the scoped permission\n     through the host; do not bypass permission checks to complete the task.\n4. Capture the response with `--output-format text` by default. Use\n   `--output-format json` only when the caller needs structured metadata.\n5. You own the result: inspect the diff, run the relevant tests, and only then\n   report it. \"Claude says it's done\" is not done.\n\n## Liveness\n\n`claude -p` is a networked, authenticated non-interactive run. If it fails,\nclassify the failure before retrying.\n\n- Auth failure: do not repair credentials unless the user asks; report the\n  blocker. If sandboxed `claude -p` says `Not logged in`, retry once with the\n  required approval because local auth can depend on keychain\u002Fsession access.\n- Sandbox\u002Fnetwork failure: rerun with the required approval if the command is\n  important to the task.\n- Long run: prefer `--max-budget-usd` or a narrower prompt before raising\n  effort.\n- Workspace trust: `-p` skips the interactive trust dialog, so run it only from\n  a repo or worktree you intentionally trust.\n\n## Rules\n\n- Use `claude -p --model opus --effort high` as the default invocation; keep\n  the alias unversioned so it follows the current Opus release.\n- Let the user override model and effort in plain language; translate that to\n  `--model` and `--effort`.\n- Keep Claude's writes out of the working tree unless delegation was explicit.\n- Do not touch the same working tree while Claude is making edits.\n- Treat Claude output as evidence to verify, not authority to relay.\n- Match the tool surface to the output contract; a claim shape you granted no\n  tool for comes back invented.\n","# Claude\n\nClaude Code is an independent agent on PATH (`claude`), sharing this working\ntree and already authenticated. Use it as a second opinion or delegated worker,\nnot as ground truth: verify its claims, own any changes, and keep its task\ncontract narrow.\n\n## Defaults\n\nUse `claude -p --model opus --effort high` unless the user or task calls for a\ndifferent model or thinking level. Keep `opus` as an alias so Claude Code\nresolves the current Opus release; do not pin a dated model ID in this skill.\n\n- **Model:** prefer aliases such as `opus`, `sonnet`, or `fable`, which follow\n  the current Claude releases. Accept a full model name only when the user or\n  task requires that specific model, and verify it is current first. Default\n  `opus`.\n- **Thinking level:** pass Claude Code's `--effort` flag. Valid levels are\n  `low`, `medium`, `high`, `xhigh`, and `max`. Default `high`.\n- **Cost guard:** for exploratory asks, add `--max-budget-usd \u003Camount>` when\n  the user gives a budget or the task is likely to sprawl.\n\nExample:\n\n```sh\nclaude -p --model opus --effort high \"\u003Cprompt>\"\n```\n\n## Prompting Claude\n\nPrompt Claude like an operator: compact, block-structured, and explicit about\nthe artifact you need. A sharper contract beats higher effort.\n\n- **One task per run.** Split review, implementation, and docs into separate\n  prompts.\n- **Name installed repo skills when relevant.** Claude can read repo files;\n  say \"follow codebase-design for module boundaries\" or \"follow write-skills\n  for the skill\" instead of pasting those skills into the prompt.\n- **Blocks, added only where useful:**\n  - `\u003Ctask>` — the concrete job, context, and expected end state.\n  - `\u003Coutput_contract>` — exact response shape, highest-value first.\n  - `\u003Cdefault_follow_through>` — take low-risk interpretations and continue;\n    stop only when a missing detail changes correctness, safety, or an\n    irreversible action.\n  - `\u003Cverification_loop>` — inspect the result against requirements and revise\n    before finalizing.\n  - `\u003Cgrounding>` — ground every claim in code or tool output; label\n    inferences.\n  - `\u003Caction_safety>` — keep the diff tightly scoped; no drive-by refactors.\n- **Anti-patterns:** vague prompts, mixed jobs, no output contract, asking it\n  to \"think harder\" instead of tightening the task, and accepting its answer\n  without checking the evidence.\n\n## Consultation\n\nUse Claude for consultation when the user asks for Claude's view or when a\nsecond-agent read from Claude is already authorized. Keep it read-oriented unless\nthe user explicitly asks it to edit.\n\n1. Ask for evidence, not vibes: require file paths, observed behavior, command\n   output, or labelled inferences.\n2. Prefer a read-only tool surface for reviews:\n   `--tools \"Read,Grep,Glob\" --permission-mode dontAsk`.\n3. **Never grant fewer tools than your output contract requires.** Read your\n   own prompt and ask which tool produces each claim you demanded. A judge that\n   cannot run a command will still answer a question posed in numbers — with\n   numbers, self-consistent and measured from nothing, and with the artifacts\n   it says it wrote. Trimming the tool surface does not make the answer\n   cautious; it makes the fabrication harder to spot. Either grant the tool or\n   stop asking for the claim. Add command execution only within the authorized\n   scope and the host’s normal approval controls; otherwise narrow the contract.\n4. Require the receipt with the number: any quantity in the response must\n   arrive with the command that produced it and that command's output. Findings\n   that fail this are struck, not discounted.\n5. Triage every finding against the code before acting. Report what Claude\n   flagged, what you accepted, and what you dismissed.\n\n## Implementation — explicit ask only\n\nDelegate implementation to Claude only when the user names Claude for the task.\nKeep delegation within the authorized task. In-scope follow-ups retain that\nauthorization; ask again only when the proposed delegation expands its scope.\n\n1. Slice the task sharply: goal, constraints, allowed scope, and verification.\n   An underspecified task stays with you until a fresh agent could not misread\n   it.\n2. Start from a clean tree or record the baseline commit so Claude's diff is\n   separable from yours. Tell Claude to keep scratch files (probes, dumps,\n   notes) out of the tree — its diff must contain only the deliverable.\n3. Choose permissions by what the task must do:\n   - Read\u002Freview: `--tools \"Read,Grep,Glob\" --permission-mode dontAsk`.\n   - Narrow edits: add edit tools and keep the prompt's allowed paths explicit.\n   - Broader work: use a dedicated worktree and the host’s normal approval\n     controls. If a required action is blocked, request the scoped permission\n     through the host; do not bypass permission checks to complete the task.\n4. Capture the response with `--output-format text` by default. Use\n   `--output-format json` only when the caller needs structured metadata.\n5. You own the result: inspect the diff, run the relevant tests, and only then\n   report it. \"Claude says it's done\" is not done.\n\n## Liveness\n\n`claude -p` is a networked, authenticated non-interactive run. If it fails,\nclassify the failure before retrying.\n\n- Auth failure: do not repair credentials unless the user asks; report the\n  blocker. If sandboxed `claude -p` says `Not logged in`, retry once with the\n  required approval because local auth can depend on keychain\u002Fsession access.\n- Sandbox\u002Fnetwork failure: rerun with the required approval if the command is\n  important to the task.\n- Long run: prefer `--max-budget-usd` or a narrower prompt before raising\n  effort.\n- Workspace trust: `-p` skips the interactive trust dialog, so run it only from\n  a repo or worktree you intentionally trust.\n\n## Rules\n\n- Use `claude -p --model opus --effort high` as the default invocation; keep\n  the alias unversioned so it follows the current Opus release.\n- Let the user override model and effort in plain language; translate that to\n  `--model` and `--effort`.\n- Keep Claude's writes out of the working tree unless delegation was explicit.\n- Do not touch the same working tree while Claude is making edits.\n- Treat Claude output as evidence to verify, not authority to relay.\n- Match the tool surface to the output contract; a claim shape you granted no\n  tool for comes back invented.\n",{"name":5,"description":7},"main","fc9ff73613f4346442defd1318eae264af415d72","skills\u002Fclaude",1,"https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit","https:\u002F\u002Ftransitivebullsh.it","My heart is open source 💕 ","2026-10-24 01:38:48",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add transitive-bullshit\u002Fskills --skill claude",{"tool":40,"command":41},"bunx","bunx skills add transitive-bullshit\u002Fskills --skill claude",{"tool":43,"command":44},"pnpm","pnpm dlx skills add transitive-bullshit\u002Fskills --skill claude","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fclaude\u002FSKILL.md",true,[48],{"path":49,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":16},"LICENSE","许可证",1068,[53,60,67,75,82,89],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":20},"01M7C8Z0JQ0VH6YYSJQGXJKVV4","xurl","Paid X API fallback through xurl after cheaper X reads fail, or for authorized account actions unsupported by bird.","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fxurl\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":20},"01M7C8YY74K2MJYK311R1PG31A","x-data","Read or research X\u002FTwitter posts, URLs, profiles, threads, or the user's X history using the cheapest available source.","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fx-data\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":71,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":20},"01M7C8YV6AAFSCTS39C2MBKTTV","write-skills","Create or revise agent skills. Use when adding a new skill file, renaming a skill, simplifying an existing skill, improving trigger descriptions, or deciding what belongs in a skill versus references, scripts, assets, or ordinary docs.","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fwrite-skills\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fwrite-skills\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":20},"01M7C8YQFGTFVQQJCH8SPFS3YQ","web-perf","Audit, diagnose, or optimize website loading and interaction performance, Core Web Vitals, and Lighthouse performance scores.","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fweb-perf\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":20},"01M7C8YJ6QFV6QFCJESHGZZ2B8","vercel-react-best-practices","Apply React and Next.js performance guidance. Use for performance work or a specific rendering, data-fetching, or bundle concern.","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-react-best-practices\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":20},"01M7C8YBRVCMAZ99D1Y5QXHMG8","vercel-optimize","Audit deployed Vercel cost and performance using project metrics. Use for expensive or slow routes, usage analysis, and caching opportunities.","https:\u002F\u002Fgithub.com\u002Ftransitive-bullshit\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-optimize\u002FSKILL.md",[],[]]