Imported from justin-delano/Claude-Config (skills/conditional/beads/audit/SKILL.md). Install upstream with npx skills add justin-delano/Claude-Config --skill audit. Copyright stays with the author.

Beads audit

Periodic maintenance and health checks for the beads issue graph. This command focuses on structural integrity of the beads graph.

When to use

Run beads audit during:

• Weekly maintenance cycles
• After completing major milestones or epics
• Before major planning sessions
• When the issue graph feels disorganized or inconsistent
• After batch imports or major graph modifications

Health check commands

Run these commands to assess graph health:

# Overall graph status
bd status

# Check and fix installation health (primary diagnostic tool)
bd doctor

# Check issues for missing template sections
bd lint

# Detect circular dependencies
bd dep cycles

# Find stale issues (not updated recently)
bd stale

# Identify orphaned issues (referenced in commits but still open)
bd orphans

# Show what's actually ready to work on
bd ready

# Show blocked issues
bd blocked

# List all issues (for manual inspection)
bd list

# Record and label agent interactions
bd audit record --prompt "..." --response "..."

Common problems

Orphaned issues

Symptoms: Issues with no epic parent, broken dependencies pointing to deleted issues, issues that should be part of an epic but aren't, or issues referenced in commits but still open.

Detection:

bd orphans                         # Identifies issues referenced in commits but still open
bd list --parent <epic-id>        # List children of specific epic
bd doctor                          # Comprehensive health check (includes broken references)
bd doctor --deep                   # Deep graph integrity validation

Remediation:

bd update <id> --parent <epic-id>  # Assign to epic (use --parent flag)
bd dep remove <id> <bad-dep-id>    # Remove broken dependency
bd delete <id>                      # Delete truly orphaned issue
bd orphans --fix                    # Close orphaned issues with confirmation
bd repair                           # Clean orphaned dependencies, labels, comments, events

Dependency cycles

Symptoms: Issues that transitively depend on themselves, blocking all progress.

Detection:

bd dep cycles

Remediation:

bd dep remove <id> <id>              # Break the cycle
bd dep add <id> <correct-dep-id>     # Rewire correctly

Stale issues

Symptoms: Issues not updated in weeks/months that may be obsolete, completed but not closed, or abandoned.

Detection:

bd stale
bd list | sort -k3              # Sort by update timestamp

Remediation:

bd delete <id>                  # Remove obsolete issue
bd update <id> --status done    # Close completed work
bd update <id> --status open    # Reset abandoned work

Status drift

Symptoms: Issues marked in_progress but no longer actively worked, blocking dependent issues.

Detection:

bd list --status in_progress | grep -v "$(date +%Y-%m-%d)"
bd stale

Remediation:

bd update <id> --status open         # Reset to open
bd update <id> --status blocked      # Mark blocked if waiting
bd close <id>                         # Close if actually complete (prefer bd close over --status done)
bd blocked                            # Show all blocked issues for review

Missing dependencies

Symptoms: Work that should block other work but doesn't, causing premature "ready" status.

Detection:

bd ready                        # Review what's marked ready
bd list --parent <epic-id>      # Review epic structure (use --parent flag)

Remediation:

bd dep add <dependent-id> <blocker-id>

Test and garbage pollution

Symptoms: Test issues, experiments, or duplicate issues cluttering the graph.

Detection:

bd doctor --check=pollution                  # Detect test/garbage issues
bd list | grep -i "test\|experiment\|tmp"    # Manual search for test patterns

Remediation:

bd doctor --check=pollution --clean          # Delete test issues (with confirmation)
bd delete <id>                                # Manually remove test issues

Graph visualization

If the bv viewer is available:

# Interactive visual exploration
bv

# Machine-readable health data
bv --robot-triage

# Check for drift from baseline
bv --check-drift

Use visualization to:

• Identify disconnected subgraphs
• Spot overly complex dependency chains
• Verify epic decomposition makes sense
• Find issues that should be merged

Remediation patterns

Restructuring an epic

When epic structure doesn't match implementation reality:

# List current structure
bd list --parent <epic-id>

# Move issues to different epic
bd update <id> --parent <new-epic-id>

# Split epic (create new epic, reassign issues)
bd create --type epic --title "Epic: New Focus Area"
bd update <story-id> --parent <new-epic-id>

# Merge epics (reassign all issues, delete empty epic)
bd list --parent <old-epic-id> | while read id; do
  bd update "$id" --parent <target-epic-id>
done
bd delete <old-epic-id>

Cleaning up completed work

After epic completion:

# Close epic and all stories
bd close <epic-id>
bd list --parent <epic-id> | while read id; do
  bd close "$id"
done

# Check for epics ready to close automatically
bd epic close-eligible

# Or delete if no historical value
bd delete --cascade <epic-id>

Fixing broken dependency chains

When dependencies are tangled:

# Remove all deps for issue
bd dep remove <id> $(bd show <id> | grep "depends:" | cut -d: -f2)

# Rebuild correct dependencies
bd dep add <id> <correct-dep-1>
bd dep add <id> <correct-dep-2>

# Verify ready status makes sense
bd ready

Audit checklist

Run through this checklist during periodic maintenance:

1. Structural integrity:

   • bd doctor passes with no errors
   • bd doctor --deep validates full graph integrity
   • bd dep cycles reports no cycles

2. Status accuracy:

   • bd ready shows only truly unblocked work
   • bd blocked shows expected blocked work
   • bd list --status in_progress contains only active work
   • bd stale shows minimal results

3. Graph cleanliness:

   • bd doctor --check=pollution finds no test issues
   • bd orphans finds no issues referenced in commits but still open
   • All issues have epic parents (unless they are epics)
   • No orphaned or disconnected subgraphs
   • bd lint passes for all issues

4. Logical coherence:

   • Epic decomposition matches current understanding
   • Dependencies reflect true blocking relationships
   • Issue titles and descriptions are current

5. Visualization check (if using bv):

   • Graph structure is comprehensible
   • No obvious visual clutter or complexity
   • Epics form coherent clusters

Automation potential

Consider scripting common audit tasks:

#!/usr/bin/env bash
# beads-health-check.sh

echo "=== Beads Health Check ==="
echo

echo "## Installation Health"
bd doctor || echo "FAILED: Doctor found issues"
echo

echo "## Dependency Cycles"
bd dep cycles && echo "OK: No cycles" || echo "FAILED: Cycles detected"
echo

echo "## Orphaned Issues"
orphan_count=$(bd orphans --json | jq length)
echo "Orphaned issues (in commits but still open): $orphan_count"
if [ "$orphan_count" -gt 0 ]; then
  echo "WARNING: Found orphaned issues"
fi
echo

echo "## Stale Issues"
stale_count=$(bd stale | wc -l)
echo "Stale issues: $stale_count"
if [ "$stale_count" -gt 10 ]; then
  echo "WARNING: High stale issue count"
fi
echo

echo "## Ready Work"
bd ready
echo

echo "## Blocked Work"
bd blocked
echo

echo "## In Progress (should be actively worked)"
bd list --status in_progress
echo

echo "## Template Compliance"
bd lint

Run this script weekly or before major planning sessions to catch drift early.

Programmatic usage

Most beads commands support global flags for scripting and automation:

# JSON output for machine parsing
bd status --json
bd ready --json
bd orphans --json
bd blocked --json

# Quiet mode (errors only)
bd doctor --quiet
bd lint --quiet

# Verbose/debug output
bd doctor --verbose
bd repair --verbose

Common scripting patterns:

# Check if any issues are ready
if [ "$(bd ready --json | jq length)" -gt 0 ]; then
  echo "Work available"
fi

# Export all open issues
bd list --status open --json > open-issues.json

# Batch operations with JSON parsing
bd list --json | jq -r '.[] | select(.priority == 0) | .id' | while read id; do
  bd update "$id" --assignee alice
done

Related commands

• beads-seed — Architecture docs to beads issues transition
• beads-evolve — Adaptive refinement during implementation
• beads — Comprehensive reference for all beads workflows and commands
• beads-orient — Session start diagnostics