/review-backlog

Post-process running log entries: prioritize, link, harmonize tags, regenerate auto-sections.

Usage

/review-backlog                 # Full review with all suggestions
/review-backlog --ideas         # Review only ideas (prioritize TBD items)
/review-backlog --risks         # Review low-confidence Process Memory items
/review-backlog --link #ID-XXX  # Find and link entries related to specific ID
/review-backlog --tags          # Harmonize tags only

Execution

Step 1: Parse Arguments

Check $ARGUMENTS for mode:

• No arguments or empty: Full Review Mode
• --ideas: Ideas Review Mode
• --risks: Risks Review Mode
• --link #ID-XXX: Link Discovery Mode
• --tags: Tag Harmonization Mode

Step 2: Load All Entries

1. Read .claude/RUNNING_LOG.md
2. Parse all entries from ## Entry Backlog section
3. Extract for each entry:
   • Entry ID
   • Type (Idea/Note, Consultation, Process Memory)
   • Description
   • Confidence/Priority
   • Status
   • Tags
   • Linked To (if present)

Store in memory for analysis.

---

Mode 1: Full Review (No Arguments)

Perform all analyses and present comprehensive review.

Analysis 1: Prioritization

Find Ideas with Priority = TBD:

For each TBD idea:

1. Analyze description keywords
2. Check if related to recent Process Memory entries (decisions, critical items)
3. Look for domain alignment (documentation, api, tooling, etc.)
4. Suggest priority: High/Med/Low with brief rationale

Output Format:

💡 Ideas Requiring Prioritization (N):

- #ID-20251222-001: Local AI-optimized docs
  → Suggested: High
  → Rationale: Aligns with knowledge-base work, mentioned in #ID-20251221-005

- #ID-20251221-003: Plugin permission system
  → Suggested: Med
  → Rationale: Dependent on architecture decisions, no immediate blockers

Analysis 2: Relationship Discovery

Find Related Entries:

For each entry, identify potential links based on:

• Shared keywords in descriptions
• Similar tags
• Temporal proximity (entries from same session)
• Causal relationships (decision → idea, consultation → implementation)

Output Format:

🔗 Suggested Links (N):

- #ID-20251222-001 ← #ID-20251221-008
  Reason: Both reference documentation workflows

- #ID-20251221-005 → #ID-20251221-003
  Reason: Decision in 005 impacts idea in 003

- #ID-20251220-012 ↔ #ID-20251220-015
  Reason: Both discuss marketplace architecture

Analysis 3: Tag Harmonization

Find Tag Inconsistencies:

1. Identify similar tags:

   • docs vs documentation
   • api vs api-design
   • anthropic vs anthropic-api

2. Count usage frequency

3. Suggest consolidation to most common variant

Output Format:

🏷️  Tag Harmonization Suggestions:

- Rename "docs" → "documentation" (4 entries affected)
- Merge "api" + "api-design" → "api-design" (3 entries)
- Merge "anthropic" + "anthropic-api" → "anthropic" (5 entries)

Analysis 4: Risk Highlighting

Find Low-Confidence Items:

• Process Memory entries with Confidence < 70%
• Status = Assumed (not yet validated)
• Critical signals (blocker, must-verify, etc.)

Output Format:

⚠️  Open Risks / Low-Confidence Items (N):

- #ID-20251221-004: Confidence 65%
  → Low confidence on validation approach
  → Status: Todo
  → Linked to: #ID-20251221-005

- #ID-20251220-010: Confidence 60%
  → Assumption about API behavior not yet validated
  → Status: Assumed

Step 3: Display Summary

🔍 Backlog Review Results
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

[Analysis 1: Prioritization]
[Analysis 2: Relationship Discovery]
[Analysis 3: Tag Harmonization]
[Analysis 4: Risk Highlighting]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Apply changes? [Y/n]

Step 4: Apply Changes (If User Confirms)

If user types "Y" or "yes":

1. Update Priorities: Edit entries to change TBD → High/Med/Low
2. Add Links: Add "Linked To" fields to suggested entries
3. Harmonize Tags: Rename/merge tags across affected entries
4. Regenerate Auto-Sections: Update High-Priority Ideas, Open Risks, Linked Insights sections
5. Update Last Updated timestamp in header

Confirm:

✅ Applied N changes to running log
   - Updated 5 priorities
   - Added 3 links
   - Harmonized 12 tag occurrences
   - Regenerated auto-sections

---

Mode 2: Ideas Review (--ideas)

Focus only on prioritizing TBD ideas.

Execution

1. Load all entries
2. Filter: Type = Idea/Note AND Priority = TBD
3. For each TBD idea:
   • Analyze description
   • Suggest priority with rationale
   • Identify potential links

Output:

💡 Ideas Review (N ideas with TBD priority)

#ID-20251222-001: Local AI-optimized docs
→ Suggested: High
→ Rationale: Aligns with knowledge-base goals
→ Potential Links: #ID-20251221-008 (documentation workflow)

#ID-20251221-003: Plugin permission system
→ Suggested: Med
→ Rationale: Dependent on architecture, no immediate need

Apply priority suggestions? [Y/n]

---

Mode 3: Risks Review (--risks)

Focus only on low-confidence Process Memory items.

Execution

1. Load all entries
2. Filter: Type = Process Memory AND (Confidence < 70% OR Status = Assumed)
3. For each risk item:
   • Display confidence level
   • Show status
   • Highlight if linked to ideas or consultations
   • Suggest validation steps

Output:

⚠️  Open Risks Review (N items)

#ID-20251221-004: Confidence 65%
→ Low confidence on validation approach
→ Status: Todo
→ Linked to: #ID-20251221-005 (decision fork)
→ Suggested Action: Test manual stub approach vs auto-detection

#ID-20251220-010: Confidence 60%
→ Assumption about API pagination
→ Status: Assumed
→ Suggested Action: Verify with actual API test

---

Mode 4: Link Discovery (--link #ID-XXX)

Find entries related to a specific entry ID.

Execution

1. Load all entries

2. Find target entry by ID

3. Analyze target entry:

   • Extract keywords from description
   • Extract tags
   • Note type and timestamp

4. Search all other entries for:

   • Shared keywords (≥ 2 words in common)
   • Shared tags (≥ 1 tag)
   • Temporal proximity (same day or adjacent days)
   • Causal language ("because of", "led to", "resulted in")

5. Rank by relevance score:

   • Shared keywords: +2 per keyword
   • Shared tags: +3 per tag
   • Same type: +1
   • Temporal proximity: +1
   • Causal language: +5

Output:

🔗 Entries Related to #ID-20251222-001

High Relevance (Score ≥ 7):
- #ID-20251221-008 (Score: 9)
  → Shares 3 keywords, 2 tags
  → Same day, both reference documentation

Medium Relevance (Score 4-6):
- #ID-20251220-015 (Score: 5)
  → Shares 2 tags, temporal proximity

Low Relevance (Score 1-3):
- #ID-20251219-003 (Score: 2)
  → Shares 1 keyword

Add "Linked To" field to #ID-20251222-001 with suggested links? [Y/n]

---

Mode 5: Tag Harmonization (--tags)

Focus only on tag consistency.

Execution

1. Load all entries

2. Extract all unique tags

3. Group similar tags:

   • Levenshtein distance < 3 edits
   • Common prefixes/suffixes
   • Semantic similarity (e.g., "doc" vs "documentation")

4. For each group:

   • Count usage frequency
   • Suggest consolidation to most common variant

Output:

🏷️  Tag Harmonization Report

Group 1: Documentation Tags
- "documentation" (8 uses) ← KEEP
- "docs" (4 uses) → Rename to "documentation"
- "doc" (1 use) → Rename to "documentation"

Group 2: API Tags
- "api-design" (5 uses) ← KEEP
- "api" (3 uses) → Rename to "api-design"

Group 3: Anthropic Tags
- "anthropic" (6 uses) ← KEEP
- "anthropic-api" (2 uses) → Rename to "anthropic"

Apply harmonization? [Y/n]

Changes: 13 tag occurrences across 10 entries

---

Auto-Section Regeneration

After any changes applied, regenerate auto-sections in RUNNING_LOG.md:

High-Priority Ideas

### 🔥 High-Priority Ideas

- #ID-20251222-001: Local AI-optimized docs (Backlog)
- #ID-20251221-012: Add WebSocket support (In Progress)

Criteria: Type = Idea/Note + Priority = High + Status ≠ Done

Open Risks / Low-Confidence Items

### ⚠️ Open Risks / Low-Confidence Items

- #ID-20251221-004: Validation approach uncertainty (65%)
- #ID-20251220-010: API pagination assumption (60%)

Criteria: Type = Process Memory + (Confidence < 60% OR Status = Assumed)

Linked Process Insights

### 🔗 Linked Process Insights

- #ID-20251221-005 → #ID-20251221-003 (Decision impacts plugin idea)
- #ID-20251220-012 ↔ #ID-20251220-015 (Marketplace architecture discussion)

Criteria: Any entry with "Linked To" field populated

---

Important Notes

• All analysis is AI-driven (relationship detection, priority suggestion, tag harmonization)
• Changes require user confirmation (Y/n prompt)
• Original entries never deleted, only enhanced
• Backup tip: User can check git diff before confirming changes
• Tag harmonization preserves semantic meaning while improving consistency

---

Execute the appropriate mode based on $ARGUMENTS.