Collab Proof

{
  "frames": {
    "technical": 0.0,
    "uncertainty": 0.0,
    "fork": 0.0,
    "ai_contribution": 0.0
  },
  "pruned": ["list of pruned frame names"],
  "intent": "FEATURE_BUILDING",
  "signal": "HIGH",
  "calibration_note": "one sentence explaining any exception rule applied"
}

## [YYYY-MM-DD] <title>

**Context**: [Frame A — what forced this choice]
**Decision**: what was chosen
**Alternatives considered**: [Frame C — road not taken]
**Reasoning**: why — prefix "inferred:" if reconstructed from context
**AI contribution**:
  - Identified: [Frame D — something developer missed]
  - Suggested: [Frame D — approach or alternative]
  - Developer-driven: [what the developer decided independently]
**Intent class**: [from Layer 02]
**Signal score**: HIGH
**Outcome**: implemented | pending | reversed

## [YYYY-MM-DD] <bug title>

**Root cause**: what actually caused the bug — the WHY, not just the what
**Symptom**: what the developer observed
**Fix**: what was changed
**Why this fix**: rationale — inferred if not stated explicitly
**Alternative fixes considered**: other approaches discussed (if any)
**AI contribution**:
  - Identified: [Frame D — did Claude spot the root cause?]
  - Suggested: [Frame D — fix approach or diagnostic step]
  - Developer-driven: [what the developer diagnosed/decided independently]
**Intent class**: BUG_FIXING
**Signal score**: HIGH
**Outcome**: fixed | workaround | deferred

# Session [YYYY-MM-DD HH:MM]

**Intent**: [class] (runner-up: [class if any])
**Signal**: HIGH
**Frames active**: A ([score]) / B ([score]) / C ([score]) / D ([score])

## What shipped
[grounded in git log]

## What was figured out
[Frame B + C — the reasoning, tradeoffs, debugging — what developers forget]

## Decisions made this session
[refs to DECISIONS.md entries]

## Where it got hard
[Frame B findings — uncertainty, reverts, EXPLORING/STUCK signals]

## AI contribution summary
[Frame D synthesis — one honest paragraph, calibrated]

## Next steps inferred
[what's obviously incomplete]

YYYY-MM-DD HH:MM | [intent] | HIGH | D:[score] | cache:[hit%]% | tok:[total] | <verb phrase> — <why it mattered>

python3 -c "
import json, sys
from pathlib import Path

projects = Path.home() / '.claude/projects'
files = sorted(projects.rglob('*.jsonl'), key=lambda f: f.stat().st_mtime, reverse=True)
if not files:
    print('no_data'); sys.exit()

with open(files[0]) as fp:
    lines = [json.loads(l) for l in fp if l.strip()]

ti = to = cr = cc = 0
turns = []
for i, line in enumerate(lines):
    if line.get('type') == 'assistant':
        u = line.get('message', {}).get('usage', {})
        if not u: continue
        inp = u.get('input_tokens', 0)
        ti += inp; to += u.get('output_tokens', 0)
        cr += u.get('cache_read_input_tokens', 0)
        cc += u.get('cache_creation_input_tokens', 0)
        prompt = ''
        for j in range(i-1, -1, -1):
            if lines[j].get('type') == 'user':
                c = lines[j].get('message', {}).get('content', '')
                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]
                break
        turns.append((inp, prompt))

total = ti + cr + cc
hit = cr / total * 100 if total else 0
print(f'input={ti} output={to} cache_read={cr} cache_create={cc} hit={hit:.0f} turns={len(turns)}')
turns.sort(reverse=True)
for idx, (tok, p) in enumerate(turns[:3]):
    print(f'top{idx+1}={tok}|{p}')
"

<div class="header">
  <div class="header-top">
    <div class="project-name">
    <span class="badge">                    <!-- intent class -->
  <div class="meta-row">                    <!-- date, branch, signal level text -->
  <div class="signal-container">
    <div class="signal-label">
    <div class="signal-track">
      <div class="signal-fill">             <!-- width % driven by signal score -->

<div class="section">                       <!-- frames -->
  <div class="section-title"> ... <span class="count">Layer 02 · ADHD tree-of-thought</span>
  <div class="frames-grid">
    <div class="frame-card">               <!-- pruned: class="frame-card pruned" -->
      <div class="frame-label">            <!-- Frame A / B / C / D -->
      <div class="frame-name">
      <div class="frame-score high|low">   <!-- score value -->

<div class="section">                       <!-- decisions — skip section if none -->
  <div class="section-title"> ... <span class="count">N recorded</span>
  <div class="decision-card">              <!-- one per DECISIONS.md entry -->
    <div class="decision-header">
      <div class="decision-title">
      <div class="decision-date">
    <div class="decision-fields">
      <div class="field-row">
        <div class="field-label">          <!-- Context / Decision / Alternatives / Reasoning -->
        <div class="field-value">
      <div class="field-row">              <!-- AI contribution row -->
        <div class="field-label">AI contribution</div>
        <div class="field-value">
          <div class="ai-block">
            <div class="ai-line ai-identified|ai-suggested|ai-developer">
              <span class="tag">IDENTIFIED|SUGGESTED|DEV-DRIVEN</span>
      <div class="field-row">              <!-- Outcome row -->
        <div class="field-label">Outcome</div>
        <div class="field-value">
          <span class="outcome-badge outcome-implemented|outcome-pending|outcome-reversed">

<div class="section">                       <!-- session narrative -->
  <div class="section-title">Session narrative</div>
  <div class="narrative-grid">
    <div class="narrative-card">           <!-- What shipped -->
    <div class="narrative-card">           <!-- What was figured out -->
    <div class="narrative-card">           <!-- Where it got hard -->
    <div class="narrative-card">           <!-- Next steps inferred -->

<div class="section">                       <!-- AI contribution summary -->
  <div class="section-title">AI contribution summary</div>
  <div class="narrative-card">             <!-- Frame D synthesis paragraph -->

<div class="section">                       <!-- token usage -->
  <div class="section-title">Token usage</div>
  <div class="narrative-card">             <!-- cache hit rate bar + top turns + optimization note -->

<div class="section">                       <!-- worklog tail -->
  <div class="section-title"> ... <span class="count">last N entries</span>
  <div class="worklog-entry">              <!-- one per recent WORKLOG line -->

<div class="footer">                        <!-- last commit hash · "Generated by collab-proof · timestamp" -->

cat > session-history/YYYY-MM-DD-HHMM-proof.html << 'HTMLEOF'
<!DOCTYPE html>
... (full HTML with inline CSS, no external resources)
HTMLEOF

YYYY-MM-DD HH:MM | [intent] | MEDIUM | D:[score] | cache:[hit%]% | tok:[total] | <verb phrase>

{
  "timestamp": "YYYY-MM-DD HH:MM:SS",
  "trigger": "pre-compact",
  "signal": "HIGH / MEDIUM / LOW",
  "frames": { "technical": 0.0, "uncertainty": 0.0, "fork": 0.0, "ai_contribution": 0.0 },
  "intent": "FEATURE_BUILDING",
  "key_moments": [
    "one-line description of the most important decision or finding so far"
  ]
}