Review Agent Mentor

Quality assurance and validation agent that audits outputs from other vault enhancement agents, ensuring consistency, correctness, and adherence to vault standards.

When to Use This Agent

Choose this agent when you need to:

• Validate the work produced by content curation, tagging, or MOC generation agents
• Run a comprehensive consistency check across all vault metadata and link structures
• Generate a unified quality scorecard comparing vault state before and after enhancements
• Spot-check a random sample of modified files to verify changes match reported actions

Consider alternatives when:

• You need to perform the actual enhancements rather than review them (use the relevant specialist agent)
• Your vault has not been processed by any other agent and there is nothing to review

Quick Start

Configuration

Copy
name: review-agent-mentor
type: agent
category: obsidian-ops-team

Example Invocation

Copy
claude agent:invoke review-agent-mentor "Review all recent agent enhancements and generate a quality report"

Example Output

=== Quality Assurance Review ===
Reviewing: 4 agent reports from last enhancement cycle

METADATA COMPLIANCE: 94.2% (1,043 of 1,107 notes)
  - 38 notes missing 'type' field
  - 26 notes with invalid date format

LINK INTEGRITY: 97.8%
  - 12 broken wikilinks detected post-enhancement
  - 4 bidirectional link mismatches

TAG STANDARDIZATION: 91.5%
  - 6 non-standard technology capitalizations remain
  - 3 tags exceed 3-level hierarchy limit

MOC COVERAGE: 88.4%
  - 7 directories still lack MOC after generation pass
  Reason: below min_notes_for_moc threshold

OVERALL VAULT HEALTH SCORE: 93.1% (up from 78.4%)

Core Concepts

Review Dimensions Overview

Aspect	Details
Metadata Compliance	Validates frontmatter fields against the vault schema definition
Link Integrity	Checks for broken, orphaned, and unidirectional link references
Tag Standardization	Verifies taxonomy adherence, proper casing, and hierarchy depth
MOC Completeness	Ensures all qualifying directories have current navigational hubs

Review Pipeline Architecture

┌─────────────┐     ┌─────────────┐
│  Agent       │────▶│  Change     │
│  Reports     │     │  Validator  │
└─────────────┘     └─────────────┘
        │                   │
        ▼                   ▼
┌─────────────┐     ┌─────────────┐
│  Spot-Check  │────▶│  Scorecard  │
│  Sampler     │     │  Generator  │
└─────────────┘     └─────────────┘

Configuration

Parameter	Type	Default	Description
sample_size	integer	25	Number of randomly selected files for spot-check validation
report_directory	string	"/System_Files/"	Directory where agent enhancement reports are stored
compliance_threshold	float	0.90	Minimum metadata compliance ratio before flagging failures
check_bidirectional	boolean	true	Whether to validate that links are reciprocated both ways
score_weights	object	{"meta":0.3,"links":0.3,"tags":0.2,"moc":0.2}	Relative weights for computing the overall vault health score

Best Practices

1. Run Reviews Immediately After Enhancement Cycles The review agent is most effective when run right after other agents complete their work. This catches issues while the context is fresh and before additional manual edits obscure what was changed by automation.

2. Use Spot-Check Sampling for Large Vaults Reviewing every file in a vault with thousands of notes is impractical. The sample_size parameter lets you statistically validate quality without exhaustive checks. A sample of 25-50 files provides high confidence for vaults under 2,000 notes.

3. Set Compliance Thresholds Realistically A 100% compliance target causes alert fatigue from edge cases like intentionally unconventional notes. Setting the threshold at 90% catches systematic problems while tolerating acceptable exceptions.

4. Track Health Scores Across Review Cycles Store each review scorecard to build a longitudinal view of vault quality. Plotting health scores over time reveals whether your maintenance practices are effective and highlights regressions early.

5. Investigate Systematic Failures, Not Individual Ones If 30 notes are missing the same frontmatter field, that is a template or process issue rather than 30 individual problems. Focus review energy on patterns that indicate upstream process failures.

Common Issues

1. Review reports reference agent output files that no longer exist If agent reports are cleaned up or moved between runs, the review agent cannot locate them for validation. Ensure report_directory is stable and consider archiving old reports rather than deleting them.

2. Spot-check sampling misses a cluster of problematic files Random sampling can miss issues concentrated in a single directory. Supplement random sampling with targeted directory checks by passing --focus-directory "path/to/suspect/folder" for known problem areas.

3. Health score appears artificially high due to weight distribution If your vault has excellent metadata but poor link integrity, default weights may mask the link problems. Adjust score_weights to emphasize the dimensions most critical to your workflow.