Kaiten Audit

Scope Controls

The user may specify audit scope:

If scope is space or board, limit Discovery and all phases to the specified entity.

Phase 0: Discovery (MANDATORY FIRST)

This phase builds the maps that all subsequent phases reference. Never skip it.

Tools

kaiten_get_company()
kaiten_get_current_user()
kaiten_list_spaces()
kaiten_list_users(include_inactive=true, limit=200)

For each space:

kaiten_list_boards(space_id=X)
kaiten_list_columns(board_id=Y)  -- for each board
kaiten_get_tree(space_id=X)

Output data structures

Build these maps in memory:

• space_map: {space_id: {name, access, boards: [...]}}
• board_map: {board_id: {name, space_id, columns: [...]}}
• column_map: {column_id: {title, type, wip_limit, wip_limit_type, archive_after_days, last_moved_warning_after_days, card_hide_after_days, board_id}}
• user_map: {user_id: {full_name, last_request_date, activated, role, created}}
• entity_tree: raw tree data per space for Phase 6

API calls estimate

Phase 1: Staleness

Identify cards that haven't moved or been updated for too long.

Tool call

kaiten_list_all_cards(
  condition=1,
  relations="none",
  fields="id,title,state,board_id,column_id,last_moved_at,column_changed_at,comment_last_added_at,updated_at,created,due_date,owner_id,responsible_id,public,share_id"
)

If scope is space, add space_id=X. If scope is board, add board_id=X.

Staleness algorithm

For each card, compute days_since_move = now - last_moved_at (fall back to column_changed_at, then updated_at, then created).

Additional flags:

• owner_id is null AND responsible_id is null = UNOWNED
• due_date < now AND state != 3 = OVERDUE

Document staleness

kaiten_list_documents(limit=200)

Flag documents where updated_at is older than 180 days as STALE_DOCUMENT.

Output

• stale_cards: list of {id, title, board, column, state, days_stale, severity}
• stale_documents: list of {uid, title, days_stale}
• staleness_score: 100 * (1 - stale_count / total_count), clamped [0, 100]

Phase 2: Public Exposure

Identify entities accessible to anyone outside the company.

Data sources

From Discovery (no additional API calls):

• Spaces where access == "for_everyone" = PUBLIC_SPACE
• Documents where public == true = PUBLIC_DOCUMENT

From Phase 1 cards:

• Cards where public == true or share_id is not null = PUBLIC_CARD

Optional: audit log check

If account has audit log access:

kaiten_list_audit_logs(categories="publication", limit=50)

Output

• public_spaces: list of {id, name}
• public_cards: list of {id, title, board}
• public_documents: list of {uid, title}
• security_score: 100 - (public_count * 5), clamped [0, 100]

Phase 3: WIP Health

Evaluate Work-In-Progress discipline across boards.

Data sources

From Discovery:

• column_map with wip_limit values

From Phase 1 cards:

• Count cards per column_id where state == 2 (in_progress)

Algorithm

For each column with type == 2 (in_progress):

1. Count active cards in this column
2. If wip_limit > 0: check if count exceeds limit
3. If wip_limit == 0 or null: flag as NO_WIP_LIMIT

Stuck cards: state == 2 AND days_since_move > 14

Output

• wip_violations: list of {board, column, limit, actual, overflow}
• columns_without_wip: list of {board, column}
• stuck_cards: list of {id, title, board, column, days_stuck}
• wip_score: weighted average of % columns within WIP and % cards not stuck

Phase 4: Activity Hotspots

Map where work is happening and where boards are dead.

Tool call

For each space:

kaiten_get_all_space_activity(
  space_id=X,
  actions="card_add,card_move,card_archive,comment_add",
  created_after=<30 days ago ISO>,
  max_pages=10,
  fields="id,action,card_id,board_id,column_id,changed"
)

Algorithm

Count events per board in the last 30 days:

Count events per user (from changed.user_id or activity actor):

Also from user_map: if last_request_date is >30 days ago, flag as DORMANT.

Input/output ratio per board:

ratio = card_add_count / max(card_archive_count, 1)

• ratio > 3.0 = INPUT_HEAVY (cards piling up)
• ratio < 0.3 = OUTPUT_HEAVY (clearing backlog)

Output

• board_activity: list of {board, space, events_30d, classification}
• user_activity: list of {user, events_30d, classification}
• dead_zones: boards with 0 events
• activity_score: 100 * (active_boards / total_boards), clamped [0, 100]

Phase 5: File Hygiene (Sampling)

Check file attachment health on a sample of cards. This is a sampling-based phase -- do NOT check every card.

Sampling strategy

1. From Phase 4, pick the top 3 most active boards
2. From Phase 1 cards, pick 10-15 cards per board (newest first)
3. Total sample: 30-50 cards

Tool call

For each sampled card:

kaiten_list_card_files(card_id=X)

Flags

Output

• file_issues: list of {card_id, card_title, issue, files}
• hygiene_file_score: 100 * (1 - issue_cards / sampled_cards)

Phase 6: Document Tree Health

Evaluate the organization of the document/entity tree.

Data sources

From Discovery: entity_tree per space (from kaiten_get_tree).

Algorithm

Traverse the tree recursively:

Output

• tree_issues: list of {space, path, issue, entity}
• hygiene_tree_score: 100 * (1 - issue_count / total_entities)
• Combined hygiene_score: average of hygiene_file_score and hygiene_tree_score

Phase 7: Kanban Metrics (Delegation)

Delegate to the kaiten-metrics skill for quantitative flow metrics.

Approach

For each space (or scoped space), collect:

• SLT median (System Lead Time)
• Cycle Time median
• Throughput per month
• WIP count
• Flow Efficiency (if time tracking is used)

Use the same kaiten_list_all_cards call pattern as described in the kaiten-metrics skill:

kaiten_list_all_cards(
  space_id=X,
  condition=2,
  relations="none",
  fields="id,title,type_id,created,first_moved_to_in_progress_at,last_moved_to_done_at,time_spent_sum,time_blocked_sum,state,condition,due_date,column_id,lane_id,board_id"
)

Output

• metrics_per_space: {space_id: {slt_median, cycle_time_median, throughput_month, wip_count}}
• Include in report as summary table

Phase 8: Automation Health

Check automation and webhook configurations.

Tool calls

For each space:

kaiten_list_automations(space_id=X)
kaiten_list_webhooks(space_id=X)
kaiten_list_incoming_webhooks(space_id=X)

Flags

Output

• automation_issues: list of {space, type, name, issue}
• Include in report as summary table

Phase 9: Entity Tree Visualization

Build an annotated ASCII tree of the workspace hierarchy.

Data sources

All from previous phases -- no additional API calls.

Annotation rules

Annotate entities in the tree with findings from earlier phases:

Example output

Space: Engineering [PUBLIC]
  Board: Sprint Board [HOT]
    Queue (wip: -)
    In Progress [!WIP 12/8]
    Review (wip: 3)
    Done
  Board: Legacy Backlog [DEAD]
    Backlog
    In Progress
    Done
  Docs:
    Architecture [STALE 14mo]
    API Reference
    Empty Folder [EMPTY]

Scoring Model

Composite health score from 5 dimensions:

composite = staleness * 0.25 + flow * 0.20 + activity * 0.20 + security * 0.20 + hygiene * 0.15

Report Template

# Kaiten Account Health Audit

**Company**: {company_name} | **Date**: {date} | **Scope**: {N} spaces, {M} boards, {K} cards

## Health Score: {score}/100 ({status})

| Dimension | Score | Status | Key Finding |
|---|---|---|---|
| Staleness | {s1}/100 | {status1} | {top_finding_1} |
| Flow/WIP | {s2}/100 | {status2} | {top_finding_2} |
| Activity | {s3}/100 | {status3} | {top_finding_3} |
| Security | {s4}/100 | {status4} | {top_finding_4} |
| Hygiene | {s5}/100 | {status5} | {top_finding_5} |

## Top 3 Actions (Quick Wins)

1. {action_1} -- expected impact: {impact_1}
2. {action_2} -- expected impact: {impact_2}
3. {action_3} -- expected impact: {impact_3}

## 1. Staleness

{stale_count} stale cards out of {total_cards} ({pct}%).
{stuck_count} STUCK cards in progress (CRITICAL).

| Card | Board | Column | State | Days Stale | Severity |
|---|---|---|---|---|---|
(max 15 rows, sorted by severity then days_stale desc)

## 2. Public Exposure

{public_count} publicly accessible entities found.

- Public spaces: {list}
- Public cards: {count} (top 5 listed)
- Public documents: {count}

## 3. WIP Health

{violation_count} WIP limit violations across {board_count} boards.

| Board | Column | WIP Limit | Actual | Overflow |
|---|---|---|---|---|
(all violations)

{no_wip_count} in-progress columns have no WIP limit set.

## 4. Activity

| Board | Space | Events (30d) | Status |
|---|---|---|---|
(sorted: hotspots first, dead zones last, max 20 rows)

Dead zones: {dead_zone_list}
Dormant users: {dormant_user_list}

## 5. File Hygiene

Sampled {sample_count} cards across {board_count} boards.
{issue_count} cards with file naming issues.

## 6. Document Tree

{tree_issue_count} issues found in document tree structure.
- Deep nesting (>4 levels): {count}
- Empty groups: {count}
- Stale documents (>6 months): {count}

## 7. Kanban Metrics Summary

| Space | SLT Median | Cycle Time Median | Throughput/mo | WIP |
|---|---|---|---|---|
(one row per space)

## 8. Automation Health

| Space | Automations | Disabled | Webhooks | Issues |
|---|---|---|---|---|
(one row per space with automations)

## 9. Entity Tree (Annotated)

(ASCII tree from Phase 9)

## Recommendations

### Quick Wins (this week)
- {recommendation_1}
- {recommendation_2}

### Planned Work (this month)
- {recommendation_3}
- {recommendation_4}

### Strategic (this quarter)
- {recommendation_5}

## Methodology

- Staleness thresholds: queued={t1}d, in_progress={t2}d, done_unarchived={t3}d
- Activity window: {period} days
- File hygiene: sampled {n} cards from top {m} active boards
- Data sources: {tool_count} MCP tool calls, {total_calls} API requests

API Call Budget

Execution Order

Phases MUST execute in order because later phases depend on earlier data:

1. Phase 0 (Discovery) -- prerequisite for all
2. Phase 1 (Staleness) -- produces card data for Phases 2, 3, 5
3. Phase 2 (Public Exposure) -- uses Phase 0 + Phase 1 data
4. Phase 3 (WIP Health) -- uses Phase 0 + Phase 1 data
5. Phase 4 (Activity) -- independent of Phase 1, but needs Phase 0
6. Phase 5 (File Hygiene) -- needs Phase 1 cards + Phase 4 top boards
7. Phase 6 (Document Tree) -- uses Phase 0 tree data
8. Phase 7 (Kanban Metrics) -- independent, uses same card fetching pattern
9. Phase 8 (Automations) -- independent, uses Phase 0 space list
10. Phase 9 (Visualization) -- aggregates all findings

Note: Phases 4, 7, 8 are independent of each other and can run after Phase 1.

ToolSearch Queries

Load MCP tools before use. Run these ToolSearch queries at the start:

ToolSearch query: "+kaiten list all cards"
ToolSearch query: "+kaiten space activity"
ToolSearch query: "+kaiten list spaces"
ToolSearch query: "+kaiten list boards"
ToolSearch query: "+kaiten list columns"
ToolSearch query: "+kaiten tree"
ToolSearch query: "+kaiten list users"
ToolSearch query: "+kaiten company"
ToolSearch query: "+kaiten list documents"
ToolSearch query: "+kaiten list card files"
ToolSearch query: "+kaiten automations"
ToolSearch query: "+kaiten webhooks"
ToolSearch query: "+kaiten audit log"

Tips

1. Reuse card data -- Phase 1 fetches all active cards once. Phases 2, 3, and 5 all use this same dataset. Never re-fetch.
2. Respect rate limits -- at 4.5 req/s, 350 calls takes ~80 seconds. Be patient.
3. Progressive reporting -- output section-by-section as each phase completes, rather than waiting for all phases to finish.
4. Skip phases if scoped -- if scope is a single board, skip Phase 4 (activity) and Phase 8 (automations) as they add limited value.
5. Cap table rows -- staleness table: max 15 rows. Activity table: max 20 rows. Always sort by severity.
6. File hygiene is sampled -- explicitly state the sample size and boards in the report. Don't claim comprehensive file analysis.
7. Handle missing data gracefully -- some fields (last_moved_at, comment_last_added_at) may be null. Fall back to updated_at, then created.
8. Delegate metrics -- for detailed kanban metrics, refer to the kaiten-metrics skill. The audit provides summary numbers only.