[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6MFHBVFXD0RX5ABV1JFDBJW":3,"skill-same:01M6MFHBVFXD0RX5ABV1JFDBJW":58},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":33,"linkOnly":34,"markdownContent":35,"markdownBody":36,"frontMatter":37,"refName":38,"commitSha":39,"skillDirectory":40,"forksCount":13,"repositoryHtmlUrl":41,"repositoryHomepageUrl":6,"ownerHtmlUrl":42,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":43,"install":44,"resources":57},"01M6MFHBVFXD0RX5ABV1JFDBJW","collab-proof",null,"Use when you want to understand what Claude contributed vs what you drove in a session. Triggers on: \u002Fcollab-proof, session retrospective, ai contribution analysis, collaboration evidence, what did claude do.","JiayiV\u002Fquestion_mixing","JiayiV","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F54466889?v=4","01M6JFJYYMRZ3QRGP9339JNQ2K","01M6JFJZ35KB0QVT50NAHPJA7G",0,"MIT","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Fpublic-02-claude-skills-main-gemini-skills-collab-proof-skill-e5bd3a65\u002FSKILL.md",[],[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32],"collab proof","proof","collab","collaboration","retrospective","analysis","session","use","evidence","do","triggers","contribution","Session Retrospective","understand","you","2026-11-10 10:12:30",false,"---\nname: \"collab-proof\"\ndescription: \"Use when you want to understand what Claude contributed vs what you drove in a session. Triggers on: \u002Fcollab-proof, session retrospective, ai contribution analysis, collaboration evidence, what did claude do.\"\nlicense: MIT\n---\n\n# collab-proof\n\nSurfaces AI collaboration evidence the developer didn't consciously record.\nVela 3-layer pipeline × ADHD 4-frame reasoning — prompt-native, zero dependencies.\n\n---\n\n## Layer 01 — Signal detection\n\nRun `git log --oneline -10` and `git diff --stat HEAD~3..HEAD` first.\n\nClassify signal level using this rubric (pick the highest that matches):\n\n**HIGH** → full artifacts (DECISIONS.md + session-history + WORKLOG + HTML)\n- New file created, OR\n- 4+ files modified, OR\n- Explicit option comparison in conversation (\"vs\", \"instead of\", \"chose X over Y\"), OR\n- Design discussion lasted 15+ exchanges, OR\n- **Bug with root cause diagnosis** — conversation contains WHY the bug happened\n  (not just \"fixed X\" but \"the bug was caused by Y because Z\")\n\n**BUG_FIXING special rule** — override file count:\nEven if only 1 file changed, classify as HIGH if the conversation contains:\n- Root cause explanation (\"the bug was...\", \"this happened because...\", \"the issue is...\")\n- Diagnosis process (\"I checked...\", \"turned out...\", \"the problem was...\")\n- Fix rationale (\"chose this approach because...\", \"instead of X, used Y because...\")\nFile count doesn't matter for bugs — a well-diagnosed single-file fix is more valuable\nthan a 10-file feature with no discussion.\n\n**MEDIUM** → WORKLOG only\n- 1–3 files modified with no root cause discussion, OR\n- Minor feature added, no tradeoffs discussed\n\n**LOW** → silence, tell user \"Routine session — nothing recorded.\"\n- No code changes, only planning\u002Fdiscussion, OR\n- Single trivial change with no context (\"change this text\", \"fix typo\", \"rename variable\")\n\nShow the user: `Signal: HIGH \u002F MEDIUM \u002F LOW — [one-line reason]`\n\n---\n\n## Layer 02 — WorkIntentClassifier\n\nRun all four frames simultaneously against conversation context + git diff.\nScore each frame 0.0–1.0 using the rubric below. Then apply pruning and classification rules.\n\n### Frame scoring rubric\n\n**Frame A — Technical** (code churn complexity)\n- `1.0` New module\u002Ffile created, complex logic added (state machine, Lua script, novel algorithm)\n- `0.5` Existing function logic modified, simple API endpoint added\n- `0.1` Typo fix, comment change, plain text edit\n\n**Frame B — Uncertainty** (developer doubt signals)\n- `1.0` Code written then fully rolled back, explicit doubt expressed (\"이게 맞나?\", \"동작 안 하네\"), `git revert`\n- `0.5` Advice sought from Claude mid-implementation, 2+ revision requests on same area\n- `0.0` Uninterrupted directive execution — developer knew exactly what to build\n\n**Frame C — Fork** (decision branch presence)\n- `1.0` Two or more alternatives explicitly compared in conversation (A vs B)\n- `0.5` No explicit comparison but tradeoff mentioned (performance vs readability)\n- `0.0` Single standard approach applied, no alternatives considered\n\n**Frame D — AI contribution** (Claude's actual impact)\n- `1.0` Claude identified a bug\u002Fedge case the developer hadn't noticed and proposed the fix\n- `0.6` Claude generated structural boilerplate\u002Fskeleton that significantly accelerated execution\n- `0.2` Claude reformatted or transcribed developer-directed code without independent contribution\n\n---\n\n### Pruning rule\n\nPrune any frame scoring \u003C 0.4.\n\n**Exception — High-Speed Execution Guard:**\nIf `Frame A >= 0.8` AND `Frame D >= 0.6`, do NOT prune and do NOT silence the session,\neven if Frame B = 0.0 and Frame C = 0.0.\nThis is a boilerplate-heavy FEATURE_BUILDING session. Classify immediately as `FEATURE_BUILDING` with `HIGH` signal.\nRationale: zero uncertainty in a fast-moving session is a feature, not a reason to discard it.\n\n---\n\n### Intent classification\n\n| Surviving frames | Dominant intent | Meaning |\n|---|---|---|\n| A high + D mid-high (B, C low) | `FEATURE_BUILDING` | High-velocity feature generation, Claude scaffolding |\n| B high + A\u002FD high | `BUG_FIXING` or `STUCK` | Active debugging or unresolved looping |\n| C high + A high | `REFACTORING` or `EXPLORING` | Architecture exploration, weighing alternatives |\n| All frames \u003C 0.4 | `FLOW_STATE` or LOW | Routine typing, silence unless Layer 01 was HIGH |\n\nIf multiple intents tie, pick the one with the highest combined frame score.\nRecord the runner-up — it belongs in the session narrative.\n\n---\n\n### Internal output format\n\nBefore proceeding to Layer 03, resolve to this structure (show it to the user):\n\n```json\n{\n  \"frames\": {\n    \"technical\": 0.0,\n    \"uncertainty\": 0.0,\n    \"fork\": 0.0,\n    \"ai_contribution\": 0.0\n  },\n  \"pruned\": [\"list of pruned frame names\"],\n  \"intent\": \"FEATURE_BUILDING\",\n  \"signal\": \"HIGH\",\n  \"calibration_note\": \"one sentence explaining any exception rule applied\"\n}\n```\n\n---\n\n## Layer 03 — Output\n\n### If HIGH signal\n\n**Append to `DECISIONS.md`** — one entry per real fork (Frame C must confirm alternatives existed):\n\n```markdown\n## [YYYY-MM-DD] \u003Ctitle>\n\n**Context**: [Frame A — what forced this choice]\n**Decision**: what was chosen\n**Alternatives considered**: [Frame C — road not taken]\n**Reasoning**: why — prefix \"inferred:\" if reconstructed from context\n**AI contribution**:\n  - Identified: [Frame D — something developer missed]\n  - Suggested: [Frame D — approach or alternative]\n  - Developer-driven: [what the developer decided independently]\n**Intent class**: [from Layer 02]\n**Signal score**: HIGH\n**Outcome**: implemented | pending | reversed\n```\n\nIf no real fork existed → write nothing. Never fabricate decisions.\n\n**BUG_FIXING intent: use this format instead:**\n\n```markdown\n## [YYYY-MM-DD] \u003Cbug title>\n\n**Root cause**: what actually caused the bug — the WHY, not just the what\n**Symptom**: what the developer observed\n**Fix**: what was changed\n**Why this fix**: rationale — inferred if not stated explicitly\n**Alternative fixes considered**: other approaches discussed (if any)\n**AI contribution**:\n  - Identified: [Frame D — did Claude spot the root cause?]\n  - Suggested: [Frame D — fix approach or diagnostic step]\n  - Developer-driven: [what the developer diagnosed\u002Fdecided independently]\n**Intent class**: BUG_FIXING\n**Signal score**: HIGH\n**Outcome**: fixed | workaround | deferred\n```\n\n**Create `session-history\u002FYYYY-MM-DD-HHMM.md`**:\n\n```markdown\n# Session [YYYY-MM-DD HH:MM]\n\n**Intent**: [class] (runner-up: [class if any])\n**Signal**: HIGH\n**Frames active**: A ([score]) \u002F B ([score]) \u002F C ([score]) \u002F D ([score])\n\n## What shipped\n[grounded in git log]\n\n## What was figured out\n[Frame B + C — the reasoning, tradeoffs, debugging — what developers forget]\n\n## Decisions made this session\n[refs to DECISIONS.md entries]\n\n## Where it got hard\n[Frame B findings — uncertainty, reverts, EXPLORING\u002FSTUCK signals]\n\n## AI contribution summary\n[Frame D synthesis — one honest paragraph, calibrated]\n\n## Next steps inferred\n[what's obviously incomplete]\n```\n\n**Append to `WORKLOG.md`**:\n```\nYYYY-MM-DD HH:MM | [intent] | HIGH | D:[score] | cache:[hit%]% | tok:[total] | \u003Cverb phrase> — \u003Cwhy it mattered>\n```\n\nFields:\n- `D:[score]` — Frame D AI contribution score (0.0–1.0)\n- `cache:[hit%]%` — cache hit rate from token analysis (or `cache:n\u002Fa` if no data)\n- `tok:[total]` — total tokens this session (input + cache_read + cache_create + output, in K e.g. `45K`)\n- verb phrase — what shipped, grounded in git log\n\n**Collect token usage** (bash — run this and capture output):\n```bash\npython3 -c \"\nimport json, sys\nfrom pathlib import Path\n\nprojects = Path.home() \u002F '.claude\u002Fprojects'\nfiles = sorted(projects.rglob('*.jsonl'), key=lambda f: f.stat().st_mtime, reverse=True)\nif not files:\n    print('no_data'); sys.exit()\n\nwith open(files[0]) as fp:\n    lines = [json.loads(l) for l in fp if l.strip()]\n\nti = to = cr = cc = 0\nturns = []\nfor i, line in enumerate(lines):\n    if line.get('type') == 'assistant':\n        u = line.get('message', {}).get('usage', {})\n        if not u: continue\n        inp = u.get('input_tokens', 0)\n        ti += inp; to += u.get('output_tokens', 0)\n        cr += u.get('cache_read_input_tokens', 0)\n        cc += u.get('cache_creation_input_tokens', 0)\n        prompt = ''\n        for j in range(i-1, -1, -1):\n            if lines[j].get('type') == 'user':\n                c = lines[j].get('message', {}).get('content', '')\n                prompt = (c if isinstance(c, str) else next((x.get('text','') for x in c if isinstance(x,dict) and x.get('type')=='text'), ''))[:80]\n                break\n        turns.append((inp, prompt))\n\ntotal = ti + cr + cc\nhit = cr \u002F total * 100 if total else 0\nprint(f'input={ti} output={to} cache_read={cr} cache_create={cc} hit={hit:.0f} turns={len(turns)}')\nturns.sort(reverse=True)\nfor idx, (tok, p) in enumerate(turns[:3]):\n    print(f'top{idx+1}={tok}|{p}')\n\"\n```\n\nParse the output and include token stats in the session narrative. Then:\n\n**Generate `session-history\u002FYYYY-MM-DD-HHMM-proof.html`** — write a self-contained HTML file. Structure and class names are fixed — do not rename or reorder sections.\n\n**Fixed CSS tokens (use exactly):**\n- Background: `#0d1117`, Card: `#161b22`, Border: `#30363d`\n- Font: `font-family: 'Courier New', monospace`\n- Frame score colors: `high` → `#3fb950`, `low` → `#f85149`, pruned → `#8b949e`\n- AI line colors: `ai-identified` → `#a371f7`, `ai-suggested` → `#d29922`, `ai-developer` → `#3fb950`\n\n**Fixed HTML structure (class names must match exactly):**\n```\n\u003Cdiv class=\"header\">\n  \u003Cdiv class=\"header-top\">\n    \u003Cdiv class=\"project-name\">\n    \u003Cspan class=\"badge\">                    \u003C!-- intent class -->\n  \u003Cdiv class=\"meta-row\">                    \u003C!-- date, branch, signal level text -->\n  \u003Cdiv class=\"signal-container\">\n    \u003Cdiv class=\"signal-label\">\n    \u003Cdiv class=\"signal-track\">\n      \u003Cdiv class=\"signal-fill\">             \u003C!-- width % driven by signal score -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- frames -->\n  \u003Cdiv class=\"section-title\"> ... \u003Cspan class=\"count\">Layer 02 · ADHD tree-of-thought\u003C\u002Fspan>\n  \u003Cdiv class=\"frames-grid\">\n    \u003Cdiv class=\"frame-card\">               \u003C!-- pruned: class=\"frame-card pruned\" -->\n      \u003Cdiv class=\"frame-label\">            \u003C!-- Frame A \u002F B \u002F C \u002F D -->\n      \u003Cdiv class=\"frame-name\">\n      \u003Cdiv class=\"frame-score high|low\">   \u003C!-- score value -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- decisions — skip section if none -->\n  \u003Cdiv class=\"section-title\"> ... \u003Cspan class=\"count\">N recorded\u003C\u002Fspan>\n  \u003Cdiv class=\"decision-card\">              \u003C!-- one per DECISIONS.md entry -->\n    \u003Cdiv class=\"decision-header\">\n      \u003Cdiv class=\"decision-title\">\n      \u003Cdiv class=\"decision-date\">\n    \u003Cdiv class=\"decision-fields\">\n      \u003Cdiv class=\"field-row\">\n        \u003Cdiv class=\"field-label\">          \u003C!-- Context \u002F Decision \u002F Alternatives \u002F Reasoning -->\n        \u003Cdiv class=\"field-value\">\n      \u003Cdiv class=\"field-row\">              \u003C!-- AI contribution row -->\n        \u003Cdiv class=\"field-label\">AI contribution\u003C\u002Fdiv>\n        \u003Cdiv class=\"field-value\">\n          \u003Cdiv class=\"ai-block\">\n            \u003Cdiv class=\"ai-line ai-identified|ai-suggested|ai-developer\">\n              \u003Cspan class=\"tag\">IDENTIFIED|SUGGESTED|DEV-DRIVEN\u003C\u002Fspan>\n      \u003Cdiv class=\"field-row\">              \u003C!-- Outcome row -->\n        \u003Cdiv class=\"field-label\">Outcome\u003C\u002Fdiv>\n        \u003Cdiv class=\"field-value\">\n          \u003Cspan class=\"outcome-badge outcome-implemented|outcome-pending|outcome-reversed\">\n\n\u003Cdiv class=\"section\">                       \u003C!-- session narrative -->\n  \u003Cdiv class=\"section-title\">Session narrative\u003C\u002Fdiv>\n  \u003Cdiv class=\"narrative-grid\">\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- What shipped -->\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- What was figured out -->\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- Where it got hard -->\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- Next steps inferred -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- AI contribution summary -->\n  \u003Cdiv class=\"section-title\">AI contribution summary\u003C\u002Fdiv>\n  \u003Cdiv class=\"narrative-card\">             \u003C!-- Frame D synthesis paragraph -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- token usage -->\n  \u003Cdiv class=\"section-title\">Token usage\u003C\u002Fdiv>\n  \u003Cdiv class=\"narrative-card\">             \u003C!-- cache hit rate bar + top turns + optimization note -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- worklog tail -->\n  \u003Cdiv class=\"section-title\"> ... \u003Cspan class=\"count\">last N entries\u003C\u002Fspan>\n  \u003Cdiv class=\"worklog-entry\">              \u003C!-- one per recent WORKLOG line -->\n\n\u003Cdiv class=\"footer\">                        \u003C!-- last commit hash · \"Generated by collab-proof · timestamp\" -->\n```\n\nWrite the HTML using bash:\n```bash\ncat > session-history\u002FYYYY-MM-DD-HHMM-proof.html \u003C\u003C 'HTMLEOF'\n\u003C!DOCTYPE html>\n... (full HTML with inline CSS, no external resources)\nHTMLEOF\n```\n\nAfter writing, show: `open session-history\u002FYYYY-MM-DD-HHMM-proof.html`\n\n---\n\n### If MEDIUM signal\n\nAppend one line to `WORKLOG.md` only:\n```\nYYYY-MM-DD HH:MM | [intent] | MEDIUM | D:[score] | cache:[hit%]% | tok:[total] | \u003Cverb phrase>\n```\n\n---\n\n### If LOW signal\n\nTell user: \"Signal: LOW — Routine session, nothing recorded.\"\n\n---\n\n## Honesty rules\n\n- Never invent decisions not in the conversation or implied by the diff\n- \"inferred:\" prefix when reasoning is reconstructed\n- Frame D must be calibrated — neither overclaim nor dismiss\n- If all frames score \u003C 0.4 → write nothing\n\n---\n\n## PreCompact snapshot (context compaction defence)\n\nWhen context compaction is about to happen (triggered by the PreCompact hook),\nrun a lightweight mid-session checkpoint before context is lost:\n\n1. Compute current Layer 01 signal level from available context\n2. Score all four frames against what's visible now\n3. Write a snapshot to `session-history\u002F.tmp-TIMESTAMP.json`:\n\n```json\n{\n  \"timestamp\": \"YYYY-MM-DD HH:MM:SS\",\n  \"trigger\": \"pre-compact\",\n  \"signal\": \"HIGH \u002F MEDIUM \u002F LOW\",\n  \"frames\": { \"technical\": 0.0, \"uncertainty\": 0.0, \"fork\": 0.0, \"ai_contribution\": 0.0 },\n  \"intent\": \"FEATURE_BUILDING\",\n  \"key_moments\": [\n    \"one-line description of the most important decision or finding so far\"\n  ]\n}\n```\n\nWhen `\u002Fcollab-proof` runs at session end:\n- Read all `session-history\u002F.tmp-*.json` files\n- Merge frame scores (take max per frame across all snapshots)\n- Combine `key_moments` arrays — these preserve tradeoff discussions that were compacted away\n- Delete `.tmp-*.json` files after merging\n","# collab-proof\n\nSurfaces AI collaboration evidence the developer didn't consciously record.\nVela 3-layer pipeline × ADHD 4-frame reasoning — prompt-native, zero dependencies.\n\n---\n\n## Layer 01 — Signal detection\n\nRun `git log --oneline -10` and `git diff --stat HEAD~3..HEAD` first.\n\nClassify signal level using this rubric (pick the highest that matches):\n\n**HIGH** → full artifacts (DECISIONS.md + session-history + WORKLOG + HTML)\n- New file created, OR\n- 4+ files modified, OR\n- Explicit option comparison in conversation (\"vs\", \"instead of\", \"chose X over Y\"), OR\n- Design discussion lasted 15+ exchanges, OR\n- **Bug with root cause diagnosis** — conversation contains WHY the bug happened\n  (not just \"fixed X\" but \"the bug was caused by Y because Z\")\n\n**BUG_FIXING special rule** — override file count:\nEven if only 1 file changed, classify as HIGH if the conversation contains:\n- Root cause explanation (\"the bug was...\", \"this happened because...\", \"the issue is...\")\n- Diagnosis process (\"I checked...\", \"turned out...\", \"the problem was...\")\n- Fix rationale (\"chose this approach because...\", \"instead of X, used Y because...\")\nFile count doesn't matter for bugs — a well-diagnosed single-file fix is more valuable\nthan a 10-file feature with no discussion.\n\n**MEDIUM** → WORKLOG only\n- 1–3 files modified with no root cause discussion, OR\n- Minor feature added, no tradeoffs discussed\n\n**LOW** → silence, tell user \"Routine session — nothing recorded.\"\n- No code changes, only planning\u002Fdiscussion, OR\n- Single trivial change with no context (\"change this text\", \"fix typo\", \"rename variable\")\n\nShow the user: `Signal: HIGH \u002F MEDIUM \u002F LOW — [one-line reason]`\n\n---\n\n## Layer 02 — WorkIntentClassifier\n\nRun all four frames simultaneously against conversation context + git diff.\nScore each frame 0.0–1.0 using the rubric below. Then apply pruning and classification rules.\n\n### Frame scoring rubric\n\n**Frame A — Technical** (code churn complexity)\n- `1.0` New module\u002Ffile created, complex logic added (state machine, Lua script, novel algorithm)\n- `0.5` Existing function logic modified, simple API endpoint added\n- `0.1` Typo fix, comment change, plain text edit\n\n**Frame B — Uncertainty** (developer doubt signals)\n- `1.0` Code written then fully rolled back, explicit doubt expressed (\"이게 맞나?\", \"동작 안 하네\"), `git revert`\n- `0.5` Advice sought from Claude mid-implementation, 2+ revision requests on same area\n- `0.0` Uninterrupted directive execution — developer knew exactly what to build\n\n**Frame C — Fork** (decision branch presence)\n- `1.0` Two or more alternatives explicitly compared in conversation (A vs B)\n- `0.5` No explicit comparison but tradeoff mentioned (performance vs readability)\n- `0.0` Single standard approach applied, no alternatives considered\n\n**Frame D — AI contribution** (Claude's actual impact)\n- `1.0` Claude identified a bug\u002Fedge case the developer hadn't noticed and proposed the fix\n- `0.6` Claude generated structural boilerplate\u002Fskeleton that significantly accelerated execution\n- `0.2` Claude reformatted or transcribed developer-directed code without independent contribution\n\n---\n\n### Pruning rule\n\nPrune any frame scoring \u003C 0.4.\n\n**Exception — High-Speed Execution Guard:**\nIf `Frame A >= 0.8` AND `Frame D >= 0.6`, do NOT prune and do NOT silence the session,\neven if Frame B = 0.0 and Frame C = 0.0.\nThis is a boilerplate-heavy FEATURE_BUILDING session. Classify immediately as `FEATURE_BUILDING` with `HIGH` signal.\nRationale: zero uncertainty in a fast-moving session is a feature, not a reason to discard it.\n\n---\n\n### Intent classification\n\n| Surviving frames | Dominant intent | Meaning |\n|---|---|---|\n| A high + D mid-high (B, C low) | `FEATURE_BUILDING` | High-velocity feature generation, Claude scaffolding |\n| B high + A\u002FD high | `BUG_FIXING` or `STUCK` | Active debugging or unresolved looping |\n| C high + A high | `REFACTORING` or `EXPLORING` | Architecture exploration, weighing alternatives |\n| All frames \u003C 0.4 | `FLOW_STATE` or LOW | Routine typing, silence unless Layer 01 was HIGH |\n\nIf multiple intents tie, pick the one with the highest combined frame score.\nRecord the runner-up — it belongs in the session narrative.\n\n---\n\n### Internal output format\n\nBefore proceeding to Layer 03, resolve to this structure (show it to the user):\n\n```json\n{\n  \"frames\": {\n    \"technical\": 0.0,\n    \"uncertainty\": 0.0,\n    \"fork\": 0.0,\n    \"ai_contribution\": 0.0\n  },\n  \"pruned\": [\"list of pruned frame names\"],\n  \"intent\": \"FEATURE_BUILDING\",\n  \"signal\": \"HIGH\",\n  \"calibration_note\": \"one sentence explaining any exception rule applied\"\n}\n```\n\n---\n\n## Layer 03 — Output\n\n### If HIGH signal\n\n**Append to `DECISIONS.md`** — one entry per real fork (Frame C must confirm alternatives existed):\n\n```markdown\n## [YYYY-MM-DD] \u003Ctitle>\n\n**Context**: [Frame A — what forced this choice]\n**Decision**: what was chosen\n**Alternatives considered**: [Frame C — road not taken]\n**Reasoning**: why — prefix \"inferred:\" if reconstructed from context\n**AI contribution**:\n  - Identified: [Frame D — something developer missed]\n  - Suggested: [Frame D — approach or alternative]\n  - Developer-driven: [what the developer decided independently]\n**Intent class**: [from Layer 02]\n**Signal score**: HIGH\n**Outcome**: implemented | pending | reversed\n```\n\nIf no real fork existed → write nothing. Never fabricate decisions.\n\n**BUG_FIXING intent: use this format instead:**\n\n```markdown\n## [YYYY-MM-DD] \u003Cbug title>\n\n**Root cause**: what actually caused the bug — the WHY, not just the what\n**Symptom**: what the developer observed\n**Fix**: what was changed\n**Why this fix**: rationale — inferred if not stated explicitly\n**Alternative fixes considered**: other approaches discussed (if any)\n**AI contribution**:\n  - Identified: [Frame D — did Claude spot the root cause?]\n  - Suggested: [Frame D — fix approach or diagnostic step]\n  - Developer-driven: [what the developer diagnosed\u002Fdecided independently]\n**Intent class**: BUG_FIXING\n**Signal score**: HIGH\n**Outcome**: fixed | workaround | deferred\n```\n\n**Create `session-history\u002FYYYY-MM-DD-HHMM.md`**:\n\n```markdown\n# Session [YYYY-MM-DD HH:MM]\n\n**Intent**: [class] (runner-up: [class if any])\n**Signal**: HIGH\n**Frames active**: A ([score]) \u002F B ([score]) \u002F C ([score]) \u002F D ([score])\n\n## What shipped\n[grounded in git log]\n\n## What was figured out\n[Frame B + C — the reasoning, tradeoffs, debugging — what developers forget]\n\n## Decisions made this session\n[refs to DECISIONS.md entries]\n\n## Where it got hard\n[Frame B findings — uncertainty, reverts, EXPLORING\u002FSTUCK signals]\n\n## AI contribution summary\n[Frame D synthesis — one honest paragraph, calibrated]\n\n## Next steps inferred\n[what's obviously incomplete]\n```\n\n**Append to `WORKLOG.md`**:\n```\nYYYY-MM-DD HH:MM | [intent] | HIGH | D:[score] | cache:[hit%]% | tok:[total] | \u003Cverb phrase> — \u003Cwhy it mattered>\n```\n\nFields:\n- `D:[score]` — Frame D AI contribution score (0.0–1.0)\n- `cache:[hit%]%` — cache hit rate from token analysis (or `cache:n\u002Fa` if no data)\n- `tok:[total]` — total tokens this session (input + cache_read + cache_create + output, in K e.g. `45K`)\n- verb phrase — what shipped, grounded in git log\n\n**Collect token usage** (bash — run this and capture output):\n```bash\npython3 -c \"\nimport json, sys\nfrom pathlib import Path\n\nprojects = Path.home() \u002F '.claude\u002Fprojects'\nfiles = sorted(projects.rglob('*.jsonl'), key=lambda f: f.stat().st_mtime, reverse=True)\nif not files:\n    print('no_data'); sys.exit()\n\nwith open(files[0]) as fp:\n    lines = [json.loads(l) for l in fp if l.strip()]\n\nti = to = cr = cc = 0\nturns = []\nfor i, line in enumerate(lines):\n    if line.get('type') == 'assistant':\n        u = line.get('message', {}).get('usage', {})\n        if not u: continue\n        inp = u.get('input_tokens', 0)\n        ti += inp; to += u.get('output_tokens', 0)\n        cr += u.get('cache_read_input_tokens', 0)\n        cc += u.get('cache_creation_input_tokens', 0)\n        prompt = ''\n        for j in range(i-1, -1, -1):\n            if lines[j].get('type') == 'user':\n                c = lines[j].get('message', {}).get('content', '')\n                prompt = (c if isinstance(c, str) else next((x.get('text','') for x in c if isinstance(x,dict) and x.get('type')=='text'), ''))[:80]\n                break\n        turns.append((inp, prompt))\n\ntotal = ti + cr + cc\nhit = cr \u002F total * 100 if total else 0\nprint(f'input={ti} output={to} cache_read={cr} cache_create={cc} hit={hit:.0f} turns={len(turns)}')\nturns.sort(reverse=True)\nfor idx, (tok, p) in enumerate(turns[:3]):\n    print(f'top{idx+1}={tok}|{p}')\n\"\n```\n\nParse the output and include token stats in the session narrative. Then:\n\n**Generate `session-history\u002FYYYY-MM-DD-HHMM-proof.html`** — write a self-contained HTML file. Structure and class names are fixed — do not rename or reorder sections.\n\n**Fixed CSS tokens (use exactly):**\n- Background: `#0d1117`, Card: `#161b22`, Border: `#30363d`\n- Font: `font-family: 'Courier New', monospace`\n- Frame score colors: `high` → `#3fb950`, `low` → `#f85149`, pruned → `#8b949e`\n- AI line colors: `ai-identified` → `#a371f7`, `ai-suggested` → `#d29922`, `ai-developer` → `#3fb950`\n\n**Fixed HTML structure (class names must match exactly):**\n```\n\u003Cdiv class=\"header\">\n  \u003Cdiv class=\"header-top\">\n    \u003Cdiv class=\"project-name\">\n    \u003Cspan class=\"badge\">                    \u003C!-- intent class -->\n  \u003Cdiv class=\"meta-row\">                    \u003C!-- date, branch, signal level text -->\n  \u003Cdiv class=\"signal-container\">\n    \u003Cdiv class=\"signal-label\">\n    \u003Cdiv class=\"signal-track\">\n      \u003Cdiv class=\"signal-fill\">             \u003C!-- width % driven by signal score -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- frames -->\n  \u003Cdiv class=\"section-title\"> ... \u003Cspan class=\"count\">Layer 02 · ADHD tree-of-thought\u003C\u002Fspan>\n  \u003Cdiv class=\"frames-grid\">\n    \u003Cdiv class=\"frame-card\">               \u003C!-- pruned: class=\"frame-card pruned\" -->\n      \u003Cdiv class=\"frame-label\">            \u003C!-- Frame A \u002F B \u002F C \u002F D -->\n      \u003Cdiv class=\"frame-name\">\n      \u003Cdiv class=\"frame-score high|low\">   \u003C!-- score value -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- decisions — skip section if none -->\n  \u003Cdiv class=\"section-title\"> ... \u003Cspan class=\"count\">N recorded\u003C\u002Fspan>\n  \u003Cdiv class=\"decision-card\">              \u003C!-- one per DECISIONS.md entry -->\n    \u003Cdiv class=\"decision-header\">\n      \u003Cdiv class=\"decision-title\">\n      \u003Cdiv class=\"decision-date\">\n    \u003Cdiv class=\"decision-fields\">\n      \u003Cdiv class=\"field-row\">\n        \u003Cdiv class=\"field-label\">          \u003C!-- Context \u002F Decision \u002F Alternatives \u002F Reasoning -->\n        \u003Cdiv class=\"field-value\">\n      \u003Cdiv class=\"field-row\">              \u003C!-- AI contribution row -->\n        \u003Cdiv class=\"field-label\">AI contribution\u003C\u002Fdiv>\n        \u003Cdiv class=\"field-value\">\n          \u003Cdiv class=\"ai-block\">\n            \u003Cdiv class=\"ai-line ai-identified|ai-suggested|ai-developer\">\n              \u003Cspan class=\"tag\">IDENTIFIED|SUGGESTED|DEV-DRIVEN\u003C\u002Fspan>\n      \u003Cdiv class=\"field-row\">              \u003C!-- Outcome row -->\n        \u003Cdiv class=\"field-label\">Outcome\u003C\u002Fdiv>\n        \u003Cdiv class=\"field-value\">\n          \u003Cspan class=\"outcome-badge outcome-implemented|outcome-pending|outcome-reversed\">\n\n\u003Cdiv class=\"section\">                       \u003C!-- session narrative -->\n  \u003Cdiv class=\"section-title\">Session narrative\u003C\u002Fdiv>\n  \u003Cdiv class=\"narrative-grid\">\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- What shipped -->\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- What was figured out -->\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- Where it got hard -->\n    \u003Cdiv class=\"narrative-card\">           \u003C!-- Next steps inferred -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- AI contribution summary -->\n  \u003Cdiv class=\"section-title\">AI contribution summary\u003C\u002Fdiv>\n  \u003Cdiv class=\"narrative-card\">             \u003C!-- Frame D synthesis paragraph -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- token usage -->\n  \u003Cdiv class=\"section-title\">Token usage\u003C\u002Fdiv>\n  \u003Cdiv class=\"narrative-card\">             \u003C!-- cache hit rate bar + top turns + optimization note -->\n\n\u003Cdiv class=\"section\">                       \u003C!-- worklog tail -->\n  \u003Cdiv class=\"section-title\"> ... \u003Cspan class=\"count\">last N entries\u003C\u002Fspan>\n  \u003Cdiv class=\"worklog-entry\">              \u003C!-- one per recent WORKLOG line -->\n\n\u003Cdiv class=\"footer\">                        \u003C!-- last commit hash · \"Generated by collab-proof · timestamp\" -->\n```\n\nWrite the HTML using bash:\n```bash\ncat > session-history\u002FYYYY-MM-DD-HHMM-proof.html \u003C\u003C 'HTMLEOF'\n\u003C!DOCTYPE html>\n... (full HTML with inline CSS, no external resources)\nHTMLEOF\n```\n\nAfter writing, show: `open session-history\u002FYYYY-MM-DD-HHMM-proof.html`\n\n---\n\n### If MEDIUM signal\n\nAppend one line to `WORKLOG.md` only:\n```\nYYYY-MM-DD HH:MM | [intent] | MEDIUM | D:[score] | cache:[hit%]% | tok:[total] | \u003Cverb phrase>\n```\n\n---\n\n### If LOW signal\n\nTell user: \"Signal: LOW — Routine session, nothing recorded.\"\n\n---\n\n## Honesty rules\n\n- Never invent decisions not in the conversation or implied by the diff\n- \"inferred:\" prefix when reasoning is reconstructed\n- Frame D must be calibrated — neither overclaim nor dismiss\n- If all frames score \u003C 0.4 → write nothing\n\n---\n\n## PreCompact snapshot (context compaction defence)\n\nWhen context compaction is about to happen (triggered by the PreCompact hook),\nrun a lightweight mid-session checkpoint before context is lost:\n\n1. Compute current Layer 01 signal level from available context\n2. Score all four frames against what's visible now\n3. Write a snapshot to `session-history\u002F.tmp-TIMESTAMP.json`:\n\n```json\n{\n  \"timestamp\": \"YYYY-MM-DD HH:MM:SS\",\n  \"trigger\": \"pre-compact\",\n  \"signal\": \"HIGH \u002F MEDIUM \u002F LOW\",\n  \"frames\": { \"technical\": 0.0, \"uncertainty\": 0.0, \"fork\": 0.0, \"ai_contribution\": 0.0 },\n  \"intent\": \"FEATURE_BUILDING\",\n  \"key_moments\": [\n    \"one-line description of the most important decision or finding so far\"\n  ]\n}\n```\n\nWhen `\u002Fcollab-proof` runs at session end:\n- Read all `session-history\u002F.tmp-*.json` files\n- Merge frame scores (take max per frame across all snapshots)\n- Combine `key_moments` arrays — these preserve tradeoff discussions that were compacted away\n- Delete `.tmp-*.json` files after merging\n",{"name":5,"license":14,"description":7},"main","068c2486ccc0f8d200a3db4d21ec137064eda3b4","skills\u002Fpublic-02-claude-skills-main-gemini-skills-collab-proof-skill-e5bd3a65","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing","https:\u002F\u002Fgithub.com\u002FJiayiV","2026-08-25 16:30:52",{"source":8,"skillName":5,"commands":45,"prompt":55,"zipAvailable":56},[46,49,52],{"tool":47,"command":48},"npx","npx skills add JiayiV\u002Fquestion_mixing --skill collab-proof",{"tool":50,"command":51},"bunx","bunx skills add JiayiV\u002Fquestion_mixing --skill collab-proof",{"tool":53,"command":54},"pnpm","pnpm dlx skills add JiayiV\u002Fquestion_mixing --skill collab-proof","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Fpublic-02-claude-skills-main-gemini-skills-collab-proof-skill-e5bd3a65\u002FSKILL.md",true,[],[59,84,112,145,173,207],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":33},"01M6NN9PJS9NHWPDG4QP93BP8F","webmcpify","Make a web app agent-ready — propose a WebMCP tool manifest, integrate, verify in a real browser, heal; unrelated code stays untouched. Use for \"webmcpify\", \"add WebMCP\", or \"expose app actions to AI agents\".","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Fwebmcpify\u002FSKILL.md",[],[61,66,67,68,69,70,71,72,73,74,75,76,77,25,78,79,80,81,82,83],"github","ai agents","web","notes","code","browser","https","actions","manifest","inventory","verify","tool","ready","unrelated","app","scope","for","webmcp",{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":88,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":33},"01M6NN6KZE5RBN6NGHN2WXV1RE","vardoger-analyze","Use when the user asks to personalize the GitHub Copilot CLI assistant, adapt Copilot to their style, use vardoger, or analyze their Copilot CLI conversation history. Reads the local session directory at `~\u002F.copilot\u002Fsession-state\u002F`, extracts recurring preferences and conventions, and writes a fenced personalization block into `~\u002F.copilot\u002Fcopilot-instructions.md`. Runs entirely on the user's machine via the local `vardoger` CLI (`pipx install vardoger`); no network calls and no uploads. Triggers: 'personalize my copilot', 'analyze my copilot history', 'tailor copilot to me', 'run vardoger', 'update my copilot instructions from history', 'make copilot learn my style'.","Apache-2.0","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Fvardoger-analyze\u002FSKILL.md",[],[92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111],"vardoger analyze","analyze","directory","cli","github copilot","copilot","run","update","learn","network","my style","install","conversation","me","local","github copilot cli","pipx","my","at","copilot instructions",{"skillNo":113,"name":114,"nameZh":6,"description":115,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":116,"categories":117,"keywords":126,"publishedAt":33},"01M6NN3CZGQP3XAJ04BM1NNYGT","transloadit-media-processing","Process media files (video, audio, images, documents) using Transloadit. Use when asked to encode video to HLS\u002FMP4, generate thumbnails, resize or watermark images, extract audio, concatenate clips, add subtitles, OCR documents, or run any media processing pipeline. Covers 86+ processing robots for file transformation at scale.","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Ftransloadit-media-processing\u002FSKILL.md",[118,123],{"categoryNo":119,"categoryName":120,"taxonomyType":121,"taxonomyTypeDesc":122},"01JBSEEDCAT000000000000016","图像与多媒体","SCENARIO","使用场景",{"categoryNo":124,"categoryName":125,"taxonomyType":121,"taxonomyTypeDesc":122},"01JBSEEDCAT000000000000023","Skill与智能体开发",[127,128,129,130,131,132,133,134,135,136,137,138,139,25,140,141,142,143,82,144],"transloadit media processing","Media","media processing","processing","MCP","mcp server","documents","subtitles","pipeline","ocr","Video","Audio","mp4","images","files","transformation","free","hls",{"skillNo":146,"name":147,"nameZh":6,"description":148,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":149,"categories":150,"keywords":154,"publishedAt":33},"01M6NN28E92F9GH4WFYQ2H1Y15","terraform-azurerm-set-diff-analyzer","Analyze Terraform plan JSON output for AzureRM Provider to distinguish between false-positive diffs (order-only changes in Set-type attributes) and actual resource changes. Use when reviewing terraform plan output for Azure resources like Application Gateway, Load Balancer, Firewall, Front Door, NSG, and other resources with Set-type attributes that cause spurious diffs due to internal ordering changes.","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Fterraform-azurerm-set-diff-analyzer\u002FSKILL.md",[151],{"categoryNo":152,"categoryName":153,"taxonomyType":121,"taxonomyTypeDesc":122},"01JBSEEDCAT000000000000011","运维与部署",[155,156,157,158,159,160,161,162,163,25,164,165,166,167,168,169,170,171,172,82],"terraform","SET","terraform azurerm set diff analyzer","diff","analyzer","diffs","azure","gateway","firewall","only","resource changes","provider","resources","load balancer","attributes","front","load","balancer",{"skillNo":174,"name":175,"nameZh":6,"description":176,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":177,"categories":178,"keywords":190,"publishedAt":33},"01M6NN1QMH2TDYWTH3XY3N5XPV","technical-job-search","Use this skill when a software engineer asks for help with job search tasks: parsing or analyzing a job description, tailoring a CV\u002Fresume, writing a cover letter, evaluating a job offer, or drafting a post-interview follow-up email. Do not activate for general career advice unrelated to an active job search action.","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Ftechnical-job-search\u002FSKILL.md",[179,182,187],{"categoryNo":180,"categoryName":181,"taxonomyType":121,"taxonomyTypeDesc":122},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":183,"categoryName":184,"taxonomyType":185,"taxonomyTypeDesc":186},"01JBSEEDCAT000000000000036","内容创作者","PROFESSION","适用职业",{"categoryNo":188,"categoryName":189,"taxonomyType":185,"taxonomyTypeDesc":186},"01JBSEEDCAT000000000000038","人力资源",[191,192,193,194,195,196,197,198,93,199,25,200,201,202,203,27,204,205,82,206],"search","Job Search","technical job search","technical","writing","tasks","email","parsing","interview","up","engineer","general","activate","task","cv","letter",{"skillNo":208,"name":209,"nameZh":6,"description":210,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":211,"categories":212,"keywords":213,"publishedAt":33},"01M6NMZFA67XYS1GS09WT42235","steno-mode","Shorthand-first response compression that cuts ~40% of response tokens while preserving technical precision and exact literals. Use when the user says \"steno mode\", \"shorthand mode\", \"compressed responses\", \"token reduction\", \"brief structured output\", or invokes \u002Fsteno. Supports four compression levels: lite, brief, court, machine. Do not trigger for requests needing polished prose such as onboarding\u002Ftutorial content, stakeholder or customer-facing copy, or teaching-focused explanations.","https:\u002F\u002Fgithub.com\u002FJiayiV\u002Fquestion_mixing\u002Fblob\u002Fmain\u002Fskills\u002Fsteno-mode\u002FSKILL.md",[],[214,215,216,217,218,194,219,25,27,220,221,222,223,224,225,226,227,228,82,229],"steno mode","mode","customer","brief","first","tutorial","teaching","tokens","as","trigger","structured output","machine","output","court","requests","response compression"]