[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77XQ2PT1ZEXT96TS3PDFSC3":3,"skill-same:01M77XQ2PT1ZEXT96TS3PDFSC3":73},{"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},"01M77XQ2PT1ZEXT96TS3PDFSC3","vs-search-threads",null,"Search Codex, Claude Code, or Cursor threads, chats, transcripts, and transcriptions, then analyze what they contain. Use to find a past session or conversation, review a transcript, audit agent performance, compare sessions, locate loops or corrections, and extract decisions, outcomes, evidence, and actionable improvements.","vltansky\u002Fvs","vltansky","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F5851280?v=4","01M77XJAE230PB81N52XJHT5A1","Vlad Tansky","01M77XJAET1CMBAZ2TWTT4ZQT8",0,"MIT","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: vs-search-threads\ndescription: \"Search Codex, Claude Code, or Cursor threads, chats, transcripts, and transcriptions, then analyze what they contain. Use to find a past session or conversation, review a transcript, audit agent performance, compare sessions, locate loops or corrections, and extract decisions, outcomes, evidence, and actionable improvements.\"\n---\n\n# Search Threads\n\nTurn one or more coding-agent conversations into an evidence-backed diagnosis.\nSupport the active conversation, attached or exported transcripts, and local Codex\nor Claude Code JSONL history. Treat Cursor's Markdown export as its portable\ntranscript contract.\n\n## Guardrails\n\n- Keep project files and transcript sources read-only unless the user explicitly\n  asks to apply findings. A cross-session report under `~\u002F.vs\u002F` is output, not a\n  source mutation.\n- Do not surface secrets, credentials, private keys, hidden reasoning, or raw\n  chain-of-thought. Paraphrase sensitive evidence and omit reasoning events.\n- Separate agent behavior, tool or environment failures, user corrections, and\n  task complexity. Similar symptoms can have different owners.\n- Label claims as **verified**, **inferred**, or **unknown**. A polished summary\n  must not conceal missing transcript sections or unavailable tool results.\n- Quote only the smallest useful evidence fragment and identify its source and\n  turn. Prefer paraphrase when a quote may contain private data.\n\n## Bound the investigation\n\nApply\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fbounded-collaboration.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fbounded-collaboration.md)\nbefore widening transcript collection. Infer the requested outcome, source\nscope, proof, artifact, authority, and stop triggers from the request. Reassess\nwhen acquisition fails at the same boundary again, the source set would expand,\nor the report cannot support its next claim. Resolve answerable facts yourself;\nask only when the user must choose the session set, authority, or outcome.\n\n## 1. Resolve scope\n\nUse explicit transcript paths, thread IDs, pasted text, or the current\nconversation first. If the user says only \"analyze this thread,\" analyze the\nactive thread. If they ask for recent history without a count, use the three most\nrecent project-matching sessions.\n\nWhen the user names a thread by content instead of by path or ID (\"the session\nwhere we fixed the auth bug\"), search the host's history for those terms, rank\ncandidates by recency and match strength, and confirm the chosen session before\nanalyzing it. Report the shortlist when more than one session plausibly matches.\n\nAsk only when choosing the wrong project, session set, or comparison baseline\nwould materially change the result.\n\n## 2. Acquire transcripts\n\nPrefer host APIs or thread-reading tools when available. Otherwise use the\nnarrowest local source that matches the request:\n\n| Host | Preferred source | Local or export fallback |\n|---|---|---|\n| Codex | Active thread or thread-read tool | `~\u002F.codex\u002Fsessions\u002F**\u002Frollout-*.jsonl`; match a supplied ID or `session_meta.payload.cwd` |\n| Claude Code | Active conversation | `~\u002F.claude\u002Fprojects\u002F\u003Cproject-slug>\u002F*.jsonl` |\n| Cursor | Active chat or `@Past Chats` | Export the chat as Markdown from History |\n\nDo not scrape Cursor's SQLite history by default. Its internal schema is not a\nportable contract; request a Markdown export if the active chat is insufficient.\nBackground or remote-agent chats may require the host's own history surface.\n\nTo locate a session by content, search the host's history directory before\nnormalizing anything, so only the matching files enter scope:\n\n```bash\nrg -l -i --glob '*.jsonl' '\u003Csearch terms>' ~\u002F.claude\u002Fprojects\u002F\u003Cproject-slug>\u002F \\\n  | xargs -r ls -t | head -5\n```\n\nPrint candidate paths and modification times, not transcript bodies.\n\nFor local JSONL or exported Markdown, normalize only the files in scope:\n\n```bash\nPROJECT_ID=$(git config --get remote.origin.url 2>\u002Fdev\u002Fnull \\\n  | sed -E 's#\\.git$##; s#.*[:\u002F]([^\u002F]+\u002F[^\u002F]+)$#\\1#; s#\u002F#-#g')\n[ -z \"$PROJECT_ID\" ] && PROJECT_ID=$(basename \"$PWD\")\nNORMALIZED_PATH=\"$HOME\u002F.vs\u002F$PROJECT_ID\u002Fthread-analysis\u002Fevidence\u002Fnormalized-$(date +%Y%m%d-%H%M%S).md\"\nnode \"\u003Cskill-dir>\u002Fscripts\u002Fnormalize-transcript.mjs\" \\\n  --output \"$NORMALIZED_PATH\" \u003Ctranscript-file> [more-files]\n```\n\nThe normalizer writes user and assistant turns to disk with stable source and\nturn labels and creates a sibling `.index.json` turn index. Stdout contains only\npath, turn count, byte count, and SHA-256 metadata. Inspect the index first, then\nretrieve only the targeted turn or short adjacent turn range needed for a claim.\nFor an unfamiliar transcript, discover candidates with a bounded search such as\n`rg -n -i --max-columns 300 --max-columns-preview '\u003Cclaim terms>' \"$NORMALIZED_PATH\" | head -50`,\nmap matching lines to the index, then slice those turns. Do not print the whole\nnormalized file.\nIt excludes system, developer, tool, and reasoning records. Read raw records only\nwhen diagnosing a specific tool failure, and summarize them rather than dumping\nthem. Follow the [disk-backed evidence contract](..\u002Fvs-internal-shared\u002Freferences\u002Fdisk-backed-evidence.md).\n\nIf a transcript cannot be accessed, say exactly which source is unavailable and\ncontinue with the evidence that is available.\n\n## 3. Build an evidence ledger\n\nFor every material claim, record:\n\n- source and turn or timestamp;\n- observed event;\n- classification: verified, inferred, or unknown;\n- owner: agent, user decision, tool\u002Fenvironment, or external dependency;\n- impact on the outcome.\n\nDeduplicate repeated status messages and compacted copies of the same turn. Treat\nsummaries as secondary evidence when original turns are present.\n\n## 4. Analyze\n\nEvaluate the conversation along these axes:\n\n1. **Intent and outcome** — what the user wanted, what was delivered, and what\n   remains incomplete.\n2. **Trajectory** — the few turning points that changed the plan or result.\n3. **Decisions and corrections** — accepted choices, rejected assumptions, and\n   durable preferences.\n4. **Execution quality** — useful actions, avoidable loops, premature claims,\n   missed verification, and unnecessary interruptions.\n5. **Failure ownership** — agent mistake versus tool, auth, sandbox, network,\n   repository, or external-state failure.\n6. **Improvements** — the smallest instruction, workflow, tooling, or product\n   changes that would prevent recurrence.\n\nUse metrics only when the transcript supports them. Useful measures include\nuser corrections, repeated commands, failed tool calls, approval waits, and\nturns between request and verified outcome. Do not invent token counts, costs,\ndurations, or success rates.\n\nFor comparisons, apply the same rubric to every transcript and explain material\ndifferences in task scope or environment before ranking agents or sessions.\n\n## 5. Report\n\nLead with the diagnosis. A short single-thread analysis stays compact in\nMarkdown in chat. A complex single-thread analysis uses HTMDX when it has three\nor more material turning points, repeated-field comparisons, or enough evidence\nand caveats to exceed one short screen. Follow the shared two-layer explanation\ncontract: chat is the TLDR and the artifact is the visual human review surface.\nOmit empty sections; for a short thread, collapse the report to Bottom line,\nFindings, and Recommended changes.\n\n```markdown\n# Thread analysis\n\n## Bottom line\n\u003Coutcome, main cause, and current state>\n\n## What happened\n- \u003Cturning point> — \u003Csource:turn> — verified|inferred\n\n## Decisions and corrections\n- \u003Cdecision or durable preference> — \u003Cevidence>\n\n## Friction\n- \u003Csymptom> -> \u003Ccause or unknown> -> \u003Cimpact>\n\n## What worked\n- \u003Ceffective behavior and why it mattered>\n\n## Recommended changes\n1. \u003Csmallest high-value change> — owner: \u003Cagent|workflow|tool|product>\n\n## Evidence gaps\n- \u003Cmissing or inaccessible evidence, or \"None material\">\n```\n\nFor a cross-session comparison, use HTMDX because the repeated rubric, metrics,\nturning points, and evidence are easier to compare visually. Resolve the project\nID and create the artifact directory. Choose a comparison slug from the session\nlabels or comparison focus, using lowercase letters, numbers, and hyphens only:\n\n```bash\nPROJECT_ID=$(git config --get remote.origin.url 2>\u002Fdev\u002Fnull \\\n  | sed -E 's#\\.git$##; s#.*[:\u002F]([^\u002F]+\u002F[^\u002F]+)$#\\1#; s#\u002F#-#g')\n[ -z \"$PROJECT_ID\" ] && PROJECT_ID=$(basename \"$PWD\")\nREPORT_DIR=\"$HOME\u002F.vs\u002F$PROJECT_ID\u002Fthread-analysis\"\nREPORT_DATE=$(date +%F)\nCOMPARISON_SLUG=\"sessions\"\nREPORT_EXTENSION=\"html\"\nREPORT_BASE=\"$REPORT_DIR\u002Fthread-comparison-$REPORT_DATE-$COMPARISON_SLUG\"\nREPORT_PATH=\"$REPORT_BASE.$REPORT_EXTENSION\"\nSUFFIX=2\nmkdir -p \"$REPORT_DIR\"\nwhile ! (set -C; : > \"$REPORT_PATH\") 2>\u002Fdev\u002Fnull; do\n  REPORT_PATH=\"$REPORT_BASE-$SUFFIX.$REPORT_EXTENSION\"\n  SUFFIX=$((SUFFIX + 1))\ndone\n```\n\nReplace the generic `sessions` slug before resolving the path. The loop appends\na numeric suffix before the extension (`-2`, `-3`, and so on) when the output\nalready exists and atomically reserves the first free path, so concurrent runs\ncannot overwrite an earlier comparison. For HTMDX, copy\n`references\u002Fthread-comparison-template.html` into the reserved `$REPORT_PATH`.\nIf population fails, report the reserved empty path instead of reusing it.\n\nEdit only the HTMDX source block. Produce one canonical artifact with no\nMarkdown twin, then return the bottom line and report path in chat. Follow the\n[shared rich-artifact contract](..\u002Fvs-internal-shared\u002Freferences\u002Frich-artifacts.md).\n\nApply the security boundary to the actual report content. Paraphrase and redact\ncredentials, secrets, PII, and sensitive evidence. A sanitized comparison still\nuses HTMDX. If the user explicitly requests Markdown, or sensitive data must\nremain and no trusted local runtime is available, set `REPORT_EXTENSION=\"md\"`,\nrun the same collision-safe path resolution, and write the comparison as\n`$REPORT_DIR\u002Fthread-comparison-$REPORT_DATE-$COMPARISON_SLUG.md` without copying\nthe HTML template.\n\nDo not write findings into rules, memories, issues, or skill files without a\nseparate explicit request.\n\n## Output style\n\nApply the [shared output style](..\u002Fvs-internal-shared\u002Freferences\u002Foutput-style.md)\nto every user-facing message.\n\n## Workflow\n\nDirect: emit **Next** only. Composed: return to caller.\n\n**Prev:** thread selection | transcript export\n**Next:** done | `\u002Fvs-retro` for reviewed learnings\n**Relevant:** `\u002Fvs-recap` | `\u002Fvs-try-skill`\n","# Search Threads\n\nTurn one or more coding-agent conversations into an evidence-backed diagnosis.\nSupport the active conversation, attached or exported transcripts, and local Codex\nor Claude Code JSONL history. Treat Cursor's Markdown export as its portable\ntranscript contract.\n\n## Guardrails\n\n- Keep project files and transcript sources read-only unless the user explicitly\n  asks to apply findings. A cross-session report under `~\u002F.vs\u002F` is output, not a\n  source mutation.\n- Do not surface secrets, credentials, private keys, hidden reasoning, or raw\n  chain-of-thought. Paraphrase sensitive evidence and omit reasoning events.\n- Separate agent behavior, tool or environment failures, user corrections, and\n  task complexity. Similar symptoms can have different owners.\n- Label claims as **verified**, **inferred**, or **unknown**. A polished summary\n  must not conceal missing transcript sections or unavailable tool results.\n- Quote only the smallest useful evidence fragment and identify its source and\n  turn. Prefer paraphrase when a quote may contain private data.\n\n## Bound the investigation\n\nApply\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fbounded-collaboration.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fbounded-collaboration.md)\nbefore widening transcript collection. Infer the requested outcome, source\nscope, proof, artifact, authority, and stop triggers from the request. Reassess\nwhen acquisition fails at the same boundary again, the source set would expand,\nor the report cannot support its next claim. Resolve answerable facts yourself;\nask only when the user must choose the session set, authority, or outcome.\n\n## 1. Resolve scope\n\nUse explicit transcript paths, thread IDs, pasted text, or the current\nconversation first. If the user says only \"analyze this thread,\" analyze the\nactive thread. If they ask for recent history without a count, use the three most\nrecent project-matching sessions.\n\nWhen the user names a thread by content instead of by path or ID (\"the session\nwhere we fixed the auth bug\"), search the host's history for those terms, rank\ncandidates by recency and match strength, and confirm the chosen session before\nanalyzing it. Report the shortlist when more than one session plausibly matches.\n\nAsk only when choosing the wrong project, session set, or comparison baseline\nwould materially change the result.\n\n## 2. Acquire transcripts\n\nPrefer host APIs or thread-reading tools when available. Otherwise use the\nnarrowest local source that matches the request:\n\n| Host | Preferred source | Local or export fallback |\n|---|---|---|\n| Codex | Active thread or thread-read tool | `~\u002F.codex\u002Fsessions\u002F**\u002Frollout-*.jsonl`; match a supplied ID or `session_meta.payload.cwd` |\n| Claude Code | Active conversation | `~\u002F.claude\u002Fprojects\u002F\u003Cproject-slug>\u002F*.jsonl` |\n| Cursor | Active chat or `@Past Chats` | Export the chat as Markdown from History |\n\nDo not scrape Cursor's SQLite history by default. Its internal schema is not a\nportable contract; request a Markdown export if the active chat is insufficient.\nBackground or remote-agent chats may require the host's own history surface.\n\nTo locate a session by content, search the host's history directory before\nnormalizing anything, so only the matching files enter scope:\n\n```bash\nrg -l -i --glob '*.jsonl' '\u003Csearch terms>' ~\u002F.claude\u002Fprojects\u002F\u003Cproject-slug>\u002F \\\n  | xargs -r ls -t | head -5\n```\n\nPrint candidate paths and modification times, not transcript bodies.\n\nFor local JSONL or exported Markdown, normalize only the files in scope:\n\n```bash\nPROJECT_ID=$(git config --get remote.origin.url 2>\u002Fdev\u002Fnull \\\n  | sed -E 's#\\.git$##; s#.*[:\u002F]([^\u002F]+\u002F[^\u002F]+)$#\\1#; s#\u002F#-#g')\n[ -z \"$PROJECT_ID\" ] && PROJECT_ID=$(basename \"$PWD\")\nNORMALIZED_PATH=\"$HOME\u002F.vs\u002F$PROJECT_ID\u002Fthread-analysis\u002Fevidence\u002Fnormalized-$(date +%Y%m%d-%H%M%S).md\"\nnode \"\u003Cskill-dir>\u002Fscripts\u002Fnormalize-transcript.mjs\" \\\n  --output \"$NORMALIZED_PATH\" \u003Ctranscript-file> [more-files]\n```\n\nThe normalizer writes user and assistant turns to disk with stable source and\nturn labels and creates a sibling `.index.json` turn index. Stdout contains only\npath, turn count, byte count, and SHA-256 metadata. Inspect the index first, then\nretrieve only the targeted turn or short adjacent turn range needed for a claim.\nFor an unfamiliar transcript, discover candidates with a bounded search such as\n`rg -n -i --max-columns 300 --max-columns-preview '\u003Cclaim terms>' \"$NORMALIZED_PATH\" | head -50`,\nmap matching lines to the index, then slice those turns. Do not print the whole\nnormalized file.\nIt excludes system, developer, tool, and reasoning records. Read raw records only\nwhen diagnosing a specific tool failure, and summarize them rather than dumping\nthem. Follow the [disk-backed evidence contract](..\u002Fvs-internal-shared\u002Freferences\u002Fdisk-backed-evidence.md).\n\nIf a transcript cannot be accessed, say exactly which source is unavailable and\ncontinue with the evidence that is available.\n\n## 3. Build an evidence ledger\n\nFor every material claim, record:\n\n- source and turn or timestamp;\n- observed event;\n- classification: verified, inferred, or unknown;\n- owner: agent, user decision, tool\u002Fenvironment, or external dependency;\n- impact on the outcome.\n\nDeduplicate repeated status messages and compacted copies of the same turn. Treat\nsummaries as secondary evidence when original turns are present.\n\n## 4. Analyze\n\nEvaluate the conversation along these axes:\n\n1. **Intent and outcome** — what the user wanted, what was delivered, and what\n   remains incomplete.\n2. **Trajectory** — the few turning points that changed the plan or result.\n3. **Decisions and corrections** — accepted choices, rejected assumptions, and\n   durable preferences.\n4. **Execution quality** — useful actions, avoidable loops, premature claims,\n   missed verification, and unnecessary interruptions.\n5. **Failure ownership** — agent mistake versus tool, auth, sandbox, network,\n   repository, or external-state failure.\n6. **Improvements** — the smallest instruction, workflow, tooling, or product\n   changes that would prevent recurrence.\n\nUse metrics only when the transcript supports them. Useful measures include\nuser corrections, repeated commands, failed tool calls, approval waits, and\nturns between request and verified outcome. Do not invent token counts, costs,\ndurations, or success rates.\n\nFor comparisons, apply the same rubric to every transcript and explain material\ndifferences in task scope or environment before ranking agents or sessions.\n\n## 5. Report\n\nLead with the diagnosis. A short single-thread analysis stays compact in\nMarkdown in chat. A complex single-thread analysis uses HTMDX when it has three\nor more material turning points, repeated-field comparisons, or enough evidence\nand caveats to exceed one short screen. Follow the shared two-layer explanation\ncontract: chat is the TLDR and the artifact is the visual human review surface.\nOmit empty sections; for a short thread, collapse the report to Bottom line,\nFindings, and Recommended changes.\n\n```markdown\n# Thread analysis\n\n## Bottom line\n\u003Coutcome, main cause, and current state>\n\n## What happened\n- \u003Cturning point> — \u003Csource:turn> — verified|inferred\n\n## Decisions and corrections\n- \u003Cdecision or durable preference> — \u003Cevidence>\n\n## Friction\n- \u003Csymptom> -> \u003Ccause or unknown> -> \u003Cimpact>\n\n## What worked\n- \u003Ceffective behavior and why it mattered>\n\n## Recommended changes\n1. \u003Csmallest high-value change> — owner: \u003Cagent|workflow|tool|product>\n\n## Evidence gaps\n- \u003Cmissing or inaccessible evidence, or \"None material\">\n```\n\nFor a cross-session comparison, use HTMDX because the repeated rubric, metrics,\nturning points, and evidence are easier to compare visually. Resolve the project\nID and create the artifact directory. Choose a comparison slug from the session\nlabels or comparison focus, using lowercase letters, numbers, and hyphens only:\n\n```bash\nPROJECT_ID=$(git config --get remote.origin.url 2>\u002Fdev\u002Fnull \\\n  | sed -E 's#\\.git$##; s#.*[:\u002F]([^\u002F]+\u002F[^\u002F]+)$#\\1#; s#\u002F#-#g')\n[ -z \"$PROJECT_ID\" ] && PROJECT_ID=$(basename \"$PWD\")\nREPORT_DIR=\"$HOME\u002F.vs\u002F$PROJECT_ID\u002Fthread-analysis\"\nREPORT_DATE=$(date +%F)\nCOMPARISON_SLUG=\"sessions\"\nREPORT_EXTENSION=\"html\"\nREPORT_BASE=\"$REPORT_DIR\u002Fthread-comparison-$REPORT_DATE-$COMPARISON_SLUG\"\nREPORT_PATH=\"$REPORT_BASE.$REPORT_EXTENSION\"\nSUFFIX=2\nmkdir -p \"$REPORT_DIR\"\nwhile ! (set -C; : > \"$REPORT_PATH\") 2>\u002Fdev\u002Fnull; do\n  REPORT_PATH=\"$REPORT_BASE-$SUFFIX.$REPORT_EXTENSION\"\n  SUFFIX=$((SUFFIX + 1))\ndone\n```\n\nReplace the generic `sessions` slug before resolving the path. The loop appends\na numeric suffix before the extension (`-2`, `-3`, and so on) when the output\nalready exists and atomically reserves the first free path, so concurrent runs\ncannot overwrite an earlier comparison. For HTMDX, copy\n`references\u002Fthread-comparison-template.html` into the reserved `$REPORT_PATH`.\nIf population fails, report the reserved empty path instead of reusing it.\n\nEdit only the HTMDX source block. Produce one canonical artifact with no\nMarkdown twin, then return the bottom line and report path in chat. Follow the\n[shared rich-artifact contract](..\u002Fvs-internal-shared\u002Freferences\u002Frich-artifacts.md).\n\nApply the security boundary to the actual report content. Paraphrase and redact\ncredentials, secrets, PII, and sensitive evidence. A sanitized comparison still\nuses HTMDX. If the user explicitly requests Markdown, or sensitive data must\nremain and no trusted local runtime is available, set `REPORT_EXTENSION=\"md\"`,\nrun the same collision-safe path resolution, and write the comparison as\n`$REPORT_DIR\u002Fthread-comparison-$REPORT_DATE-$COMPARISON_SLUG.md` without copying\nthe HTML template.\n\nDo not write findings into rules, memories, issues, or skill files without a\nseparate explicit request.\n\n## Output style\n\nApply the [shared output style](..\u002Fvs-internal-shared\u002Freferences\u002Foutput-style.md)\nto every user-facing message.\n\n## Workflow\n\nDirect: emit **Next** only. Composed: return to caller.\n\n**Prev:** thread selection | transcript export\n**Next:** done | `\u002Fvs-retro` for reviewed learnings\n**Relevant:** `\u002Fvs-recap` | `\u002Fvs-try-skill`\n",{"name":5,"description":7},"master","72448991571c9e5f0e76cccf189b8522d442e6eb","skills\u002Fvs-search-threads",1,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs","https:\u002F\u002Fgithub.com\u002Fvltansky","2026-11-06 18:47:54",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add vltansky\u002Fvs --skill vs-search-threads",{"tool":37,"command":38},"bunx","bunx skills add vltansky\u002Fvs --skill vs-search-threads",{"tool":40,"command":41},"pnpm","pnpm dlx skills add vltansky\u002Fvs --skill vs-search-threads","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002FSKILL.md",true,[45,51,55,59,65,69],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"agents\u002Fopenai.yaml","REFERENCE","参考文档",169,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002Fagents\u002Fopenai.yaml",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"README.md",826,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002FREADME.md",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Fthread-comparison-template.html",1677,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002Freferences\u002Fthread-comparison-template.html",{"path":60,"resourceType":61,"resourceTypeDesc":62,"sizeBytes":63,"sourceHtmlUrl":64},"scripts\u002Fnormalize-transcript.mjs","SCRIPT","脚本",6142,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002Fscripts\u002Fnormalize-transcript.mjs",{"path":66,"resourceType":61,"resourceTypeDesc":62,"sizeBytes":67,"sourceHtmlUrl":68},"test\u002Fdisk-backed-transcript.static.eval.ts",2040,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002Ftest\u002Fdisk-backed-transcript.static.eval.ts",{"path":70,"resourceType":61,"resourceTypeDesc":62,"sizeBytes":71,"sourceHtmlUrl":72},"test\u002Frich-report.static.eval.ts",1900,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-search-threads\u002Ftest\u002Frich-report.static.eval.ts",[74,81,87,92,99,104],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M77XSKWFMPKFXM0311QKEJHH","vs-write","Use when asked to write, rewrite, edit, tighten, simplify, or polish prose for clarity, or to put text in STE, ASD-STE100, or controlled language. Produces plain, direct, accessible copy without losing substance or precision.","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-write\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"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":84,"categories":85,"keywords":86,"publishedAt":19},"01M77XSJNX2V560HSKQKQD67KP","vs-fake-verify","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fstub-rejector\u002FSKILL.md",[],[],{"skillNo":88,"name":83,"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":89,"categories":90,"keywords":91,"publishedAt":19},"01M77XSHHBEMMYZ6B4PD384NZH","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fstructure-paste\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":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":19},"01M77XSE2Y7MW0KP3DTFNEE2E8","vs-verify","Use when asked to verify, prove it works, check completion, or before claiming done. Produces evidence-backed PASS\u002FWARN\u002FFAIL status.","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fpublished-pair\u002FSKILL.md",[],[],{"skillNo":100,"name":83,"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":101,"categories":102,"keywords":103,"publishedAt":19},"01M77XSAXFQXNE6VCRC7275Z0X","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fphrase-complete-beside\u002FSKILL.md",[],[],{"skillNo":105,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":19},"01M77XS6R0E9X1E5KKSS00WYSS","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002FSKILL.md",[],[]]