[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M780NR1TDA4X8V3YCAMJFYY2":3,"skill-same:01M780NR1TDA4X8V3YCAMJFYY2":48},{"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":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M780NR1TDA4X8V3YCAMJFYY2","debug-session",null,"Hypothesis-driven debugging for bugs, with structured trace logging under .claude\u002Fdebug\u002F. Use when a bug's cause is non-obvious and print statements are about to be sprinkled in.","DTCurrie\u002Fhouserules","DTCurrie","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1930371?v=4","01M780ND83PAY17A9HX96HHWG4","Devin T. Currie","01M780ND8W21FE1NSPAG03M0HM",2,"MIT","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fcli\u002Fpayload\u002Fskills\u002Fdebug-session\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: debug-session\ndescription: Hypothesis-driven debugging for bugs, with structured trace logging under .claude\u002Fdebug\u002F. Use when a bug's cause is non-obvious and print statements are about to be sprinkled in.\nargument-hint: \u003Cbug description>\nallowed-tools: Read, Edit, Write, Grep, Glob, Bash\n---\n\nSystematically debug: **$ARGUMENTS**\n\nThree things make this cheaper than scattershot `console.log`. Hypotheses come first, every trace\nline is machine-readable and tagged, and **every edit you make is removed at the end**. The tag\n`CLAUDE-DEBUG` is the contract. It goes on every line you touch, so cleanup is a mechanical grep\nrather than a memory exercise.\n\n## 1. Frame the investigation\n\n- Restate the bug in one line: the observed behavior, the expected behavior, and the exact trigger\n  (route, command, input) that reproduces it.\n- Write **1–3 explicit, falsifiable hypotheses**, labeled `H1`, `H2`, `H3`. Each must name a\n  suspected cause _and_ the observation that would confirm or reject it. \"Something's off in auth\"\n  is not a hypothesis. \"H1: the session token is null by the time `requireUser` runs\" is.\n- Say which hypothesis you'll test first and why. Don't instrument everything at once.\n\n## 2. Open the session log\n\nOne JSONL file per investigation: `.claude\u002Fdebug\u002F\u003Cslug>.jsonl`, where `\u003Cslug>` is a short kebab of\nthe bug, such as `login-500`. Create the dir if needed:\n\n```\nmkdir -p .claude\u002Fdebug\n```\n\n`.claude\u002Fdebug\u002F` is self-gitignored, so logs never enter a commit. Write one JSON object per line.\nAlways include `hyp` (which hypothesis) and `at` (a trace-point label), plus whatever you're\nobserving:\n\n```\n{\"hyp\":\"H1\",\"at\":\"requireUser.entry\",\"token\":null,\"path\":\"\u002Fdashboard\"}\n{\"hyp\":\"H1\",\"at\":\"session.load\",\"found\":false,\"sid\":\"abc\"}\n```\n\n## 3. Instrument to test the current hypothesis\n\nAdd trace lines that emit to the session log. **Tag every added line, and any added import or\nhelper, with `CLAUDE-DEBUG`.** That tag is what makes step 9 complete and verifiable.\n\nFor JS and TS, inline the fs\u002Fjson call so the whole emit is one tagged line to delete:\n\n\u003C!-- prettier-ignore -->\n```js\nrequire('node:fs').appendFileSync('.claude\u002Fdebug\u002Flogin-500.jsonl', JSON.stringify({hyp:'H1', at:'requireUser.entry', token})+'\\n'); \u002F\u002F CLAUDE-DEBUG\n```\n\nPython:\n\n```python\nopen('.claude\u002Fdebug\u002Flogin-500.jsonl','a').write(__import__('json').dumps({'hyp':'H1','at':'require_user.entry','token':token})+'\\n')  # CLAUDE-DEBUG\n```\n\nIn other languages, write one self-contained tagged line per emit that appends a JSON object and a\nnewline to the same file.\n\nRules that keep this clean:\n\n- Keep each emit to one self-contained tagged line, so removal is a single-line delete. If you must\n  add an import or helper instead of inlining, tag it `CLAUDE-DEBUG` too.\n- The path is relative to the repo root. If the process runs from another CWD, use the absolute repo\n  path instead.\n- Instrument the _decision points_ your hypothesis is about, such as entry, branch taken, or value\n  at a boundary. Not every line. Log the values that would distinguish confirm from reject.\n\n## 4. Reproduce\n\nRun the exact trigger to emit fresh logs: start the server and hit the route, run the failing test,\nor invoke the command. Truncate the log between runs if a clean slate helps:\n`: > .claude\u002Fdebug\u002F\u003Cslug>.jsonl`.\n\n## 5. Read the evidence with `jq`\n\nQuery the log instead of eyeballing it. That's why it's structured.\n\n```\njq -c 'select(.hyp==\"H1\")' .claude\u002Fdebug\u002F\u003Cslug>.jsonl        # all H1 evidence\njq -r '.hyp' .claude\u002Fdebug\u002F\u003Cslug>.jsonl | sort | uniq -c     # how much fired, per hypothesis\njq -c 'select(.hyp==\"H1\") | {at, token, found}' .claude\u002Fdebug\u002F\u003Cslug>.jsonl   # just the deciding fields\ntail -n 20 .claude\u002Fdebug\u002F\u003Cslug>.jsonl | jq -c .              # the last events before the failure\n```\n\nFor each hypothesis, state a verdict from the output: **confirmed**, **rejected**, or **supported,\nnot conclusive**. Quote the log line that decides it.\n\n## 6. Loop\n\n- Rejected a hypothesis? Move to the next, or form a new one from what the evidence revealed.\n- Confirmed the mechanism but not the root cause? Push instrumentation one level deeper, up the call\n  stack, into the dependency, or earlier in the lifecycle, and repeat from step 3.\n- Continue until you've found a **root cause** or resolved every hypothesis. If all are rejected, say\n  so plainly and propose the next hypotheses rather than guessing at a fix.\n\n## 7. Review with the user\n\nBefore touching a fix, report the root cause, the evidence with the deciding log lines quoted, and\nwhich hypotheses were confirmed or rejected. Let the user react. They may know context that reframes\nit.\n\n## 8. Propose the fix\n\nDescribe the minimal fix and **confirm before applying it.** The fix is real source. The\ninstrumentation is not. Keep them separate, and don't smuggle a fix into the debugging edits.\n\n## 9. Remove ALL instrumentation\n\nOnce the user confirms the issue is resolved, or asks you to clean up, leave nothing orphaned:\n\n1. Find every tagged edit:\n   ```\n   git grep -n --untracked CLAUDE-DEBUG -- ':!.claude'\n   ```\n2. Remove each one. Delete tagged lines, restore any line you modified, and delete any helper file\n   you created solely for tracing.\n3. **Verify zero remain.** This is the acceptance check, and the command must print nothing:\n   ```\n   git grep -n --untracked CLAUDE-DEBUG -- ':!.claude'\n   ```\n4. Clear the session log with `rm .claude\u002Fdebug\u002F\u003Cslug>.jsonl`, leaving `.claude\u002Fdebug\u002F.gitignore` in\n   place.\n\nThe `.claude\u002Fdebug\u002F` directory should only ever hold _active_ investigations. If you're pausing an\ninvestigation to resume later, leave the log and the tagged instrumentation in place. The\nSessionStart reminder will flag both next session, so nothing is silently shipped.\n","Systematically debug: **$ARGUMENTS**\n\nThree things make this cheaper than scattershot `console.log`. Hypotheses come first, every trace\nline is machine-readable and tagged, and **every edit you make is removed at the end**. The tag\n`CLAUDE-DEBUG` is the contract. It goes on every line you touch, so cleanup is a mechanical grep\nrather than a memory exercise.\n\n## 1. Frame the investigation\n\n- Restate the bug in one line: the observed behavior, the expected behavior, and the exact trigger\n  (route, command, input) that reproduces it.\n- Write **1–3 explicit, falsifiable hypotheses**, labeled `H1`, `H2`, `H3`. Each must name a\n  suspected cause _and_ the observation that would confirm or reject it. \"Something's off in auth\"\n  is not a hypothesis. \"H1: the session token is null by the time `requireUser` runs\" is.\n- Say which hypothesis you'll test first and why. Don't instrument everything at once.\n\n## 2. Open the session log\n\nOne JSONL file per investigation: `.claude\u002Fdebug\u002F\u003Cslug>.jsonl`, where `\u003Cslug>` is a short kebab of\nthe bug, such as `login-500`. Create the dir if needed:\n\n```\nmkdir -p .claude\u002Fdebug\n```\n\n`.claude\u002Fdebug\u002F` is self-gitignored, so logs never enter a commit. Write one JSON object per line.\nAlways include `hyp` (which hypothesis) and `at` (a trace-point label), plus whatever you're\nobserving:\n\n```\n{\"hyp\":\"H1\",\"at\":\"requireUser.entry\",\"token\":null,\"path\":\"\u002Fdashboard\"}\n{\"hyp\":\"H1\",\"at\":\"session.load\",\"found\":false,\"sid\":\"abc\"}\n```\n\n## 3. Instrument to test the current hypothesis\n\nAdd trace lines that emit to the session log. **Tag every added line, and any added import or\nhelper, with `CLAUDE-DEBUG`.** That tag is what makes step 9 complete and verifiable.\n\nFor JS and TS, inline the fs\u002Fjson call so the whole emit is one tagged line to delete:\n\n\u003C!-- prettier-ignore -->\n```js\nrequire('node:fs').appendFileSync('.claude\u002Fdebug\u002Flogin-500.jsonl', JSON.stringify({hyp:'H1', at:'requireUser.entry', token})+'\\n'); \u002F\u002F CLAUDE-DEBUG\n```\n\nPython:\n\n```python\nopen('.claude\u002Fdebug\u002Flogin-500.jsonl','a').write(__import__('json').dumps({'hyp':'H1','at':'require_user.entry','token':token})+'\\n')  # CLAUDE-DEBUG\n```\n\nIn other languages, write one self-contained tagged line per emit that appends a JSON object and a\nnewline to the same file.\n\nRules that keep this clean:\n\n- Keep each emit to one self-contained tagged line, so removal is a single-line delete. If you must\n  add an import or helper instead of inlining, tag it `CLAUDE-DEBUG` too.\n- The path is relative to the repo root. If the process runs from another CWD, use the absolute repo\n  path instead.\n- Instrument the _decision points_ your hypothesis is about, such as entry, branch taken, or value\n  at a boundary. Not every line. Log the values that would distinguish confirm from reject.\n\n## 4. Reproduce\n\nRun the exact trigger to emit fresh logs: start the server and hit the route, run the failing test,\nor invoke the command. Truncate the log between runs if a clean slate helps:\n`: > .claude\u002Fdebug\u002F\u003Cslug>.jsonl`.\n\n## 5. Read the evidence with `jq`\n\nQuery the log instead of eyeballing it. That's why it's structured.\n\n```\njq -c 'select(.hyp==\"H1\")' .claude\u002Fdebug\u002F\u003Cslug>.jsonl        # all H1 evidence\njq -r '.hyp' .claude\u002Fdebug\u002F\u003Cslug>.jsonl | sort | uniq -c     # how much fired, per hypothesis\njq -c 'select(.hyp==\"H1\") | {at, token, found}' .claude\u002Fdebug\u002F\u003Cslug>.jsonl   # just the deciding fields\ntail -n 20 .claude\u002Fdebug\u002F\u003Cslug>.jsonl | jq -c .              # the last events before the failure\n```\n\nFor each hypothesis, state a verdict from the output: **confirmed**, **rejected**, or **supported,\nnot conclusive**. Quote the log line that decides it.\n\n## 6. Loop\n\n- Rejected a hypothesis? Move to the next, or form a new one from what the evidence revealed.\n- Confirmed the mechanism but not the root cause? Push instrumentation one level deeper, up the call\n  stack, into the dependency, or earlier in the lifecycle, and repeat from step 3.\n- Continue until you've found a **root cause** or resolved every hypothesis. If all are rejected, say\n  so plainly and propose the next hypotheses rather than guessing at a fix.\n\n## 7. Review with the user\n\nBefore touching a fix, report the root cause, the evidence with the deciding log lines quoted, and\nwhich hypotheses were confirmed or rejected. Let the user react. They may know context that reframes\nit.\n\n## 8. Propose the fix\n\nDescribe the minimal fix and **confirm before applying it.** The fix is real source. The\ninstrumentation is not. Keep them separate, and don't smuggle a fix into the debugging edits.\n\n## 9. Remove ALL instrumentation\n\nOnce the user confirms the issue is resolved, or asks you to clean up, leave nothing orphaned:\n\n1. Find every tagged edit:\n   ```\n   git grep -n --untracked CLAUDE-DEBUG -- ':!.claude'\n   ```\n2. Remove each one. Delete tagged lines, restore any line you modified, and delete any helper file\n   you created solely for tracing.\n3. **Verify zero remain.** This is the acceptance check, and the command must print nothing:\n   ```\n   git grep -n --untracked CLAUDE-DEBUG -- ':!.claude'\n   ```\n4. Clear the session log with `rm .claude\u002Fdebug\u002F\u003Cslug>.jsonl`, leaving `.claude\u002Fdebug\u002F.gitignore` in\n   place.\n\nThe `.claude\u002Fdebug\u002F` directory should only ever hold _active_ investigations. If you're pausing an\ninvestigation to resume later, leave the log and the tagged instrumentation in place. The\nSessionStart reminder will flag both next session, so nothing is silently shipped.\n",{"name":5,"description":7,"allowed-tools":24,"argument-hint":25},"Read, Edit, Write, Grep, Glob, Bash","\u003Cbug description>","main","039f153ef841e9e6f26430b3e17ea428d8a3ce06","packages\u002Fcli\u002Fpayload\u002Fskills\u002Fdebug-session",0,"https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules","https:\u002F\u002Fgithub.com\u002FDTCurrie","Engineering Lead at @viamrobotics, specializing in 2D and 3D visualizations.","2026-11-13 21:14:33",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add DTCurrie\u002Fhouserules --skill debug-session",{"tool":40,"command":41},"bunx","bunx skills add DTCurrie\u002Fhouserules --skill debug-session",{"tool":43,"command":44},"pnpm","pnpm dlx skills add DTCurrie\u002Fhouserules --skill debug-session","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fcli\u002Fpayload\u002Fskills\u002Fdebug-session\u002FSKILL.md",true,[],[49,56,63,70,76,83],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M780QVJX363SHD100WRD4TD0","pr-description","Write the pull request description for a change as a pasteable markdown body. Use before opening or updating a PR, or when asked for a PR body, PR summary, or gh pr create text.","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fplugin-prose\u002Fpayload\u002Fskills\u002Fpr-description\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M780QR0V9D4AJHTPN1NJHQRG","ledger-sync","Push the backlog and decision ledgers to GitHub Projects (issues and project board items). Use when the user asks to sync, push, or upload to GitHub Projects or the project board.","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fplugin-github\u002Fpayload\u002Fskills\u002Fledger-sync\u002FSKILL.md",[],[],{"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":19},"01M780QKAY9PW98XRYGQFNDEMP","backlog-adopt","Adopt a reported GitHub issue into the backlog ledger and the linked project board. Use to triage or track a GitHub issue or bug report as a backlog entry.","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fplugin-github\u002Fpayload\u002Fskills\u002Fbacklog-adopt\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M780QJ4B917TCZM76X55NKQE","design-review","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fplugin-design\u002Fpayload\u002Fskills\u002Fdesign-review\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M780QEV3HGXEWMWG5VJEFF0J","chrome-devtools-mode","Switch the Chrome DevTools MCP server between its full 29-tool surface and the slim 3-tool one (--slim). Use when asked to change, shrink, or restore the chrome-devtools MCP tool surface.","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fplugin-design\u002Fpayload\u002Fskills\u002Fchrome-devtools-mode\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M780Q9MX9ADN41RZKQ47B8CT","decide","Record a decision to the decision ledger, with the rejected alternative and revisit trigger. Use when a discussion settles a design question or a new decision supersedes a prior one.","https:\u002F\u002Fgithub.com\u002FDTCurrie\u002Fhouserules\u002Fblob\u002Fmain\u002Fpackages\u002Fplugin-decisions\u002Fpayload\u002Fskills\u002Fdecide\u002FSKILL.md",[],[]]