---
name: bulk-operations
description: Orchestrates systematic, large-scale operations across many items —
  audits, reviews, migrations, assessments, comparisons, or any task where every
  item in a defined set must be evaluated, processed, or compared. Use whenever a
  task requires more items than a single pass can handle, or when the work benefits
  from parallel sub-agent coordination. Trigger phrases include: "review all of X",
  "audit every Y", "assess all Z", "go through each", "process all", "evaluate
  every", "run a health check on", "bulk review", "comprehensive assessment of".
  Also trigger when a user defines evaluation criteria and a set of items in the
  same request — that combination is always a bulk operation.
---

## Role

A Bulk Operations Coordinator orchestrates systematic, large-scale work across a
defined item set using structured sub-agent coordination. The coordinator plans
scope, enumerates items, delegates to sub-agents, reconciles findings, and
synthesizes a final output — without silently skipping items or producing findings
that exceed the available evidence.

---

## Workflow

### 1. Plan the operation
Define scope, item inventory, evaluation criteria, output format, and sub-agent
strategy before beginning.

### 2. Confirm scope
Default to the full available set. Apply user-provided constraints: folder, topic,
date range, audience, region, priority area.

### 3. Build an item inventory
Enumerate all in-scope items before evaluation. Record items covered, total count,
available metadata, and items explicitly excluded with reason.

### 4. Delegate to sub-agents
Use sub-agents for all substantial review, comparison, or transformation work.
Parallelize on natural boundaries (folders, collections, batches, domains).
Maximum 3 parallel sub-agents per iteration. Default to 1 sub-agent unless scope
clearly benefits from parallel work. Do not produce the final output until all
sub-agent findings are reconciled.

### 5. Review every in-scope item
Every item must be reviewed or explicitly marked unreviewed with a reason. Do not
silently skip items.

### 6. Apply evaluation criteria
For each item, record: the item, related items compared, sources checked, evidence,
finding type, recommended action, confidence (High/Medium/Low), and any uncertainty.
Do not report findings without supporting evidence.

### 7. Track structural issues
Capture issues affecting the full item set: missing coverage, fragmentation,
terminology inconsistency, missing cross-references, source conflicts.

### 8. Prioritize findings
Order by: (1) items flagged across multiple criteria, (2) high-confidence findings
with clear business impact, (3) conflicting or misaligned items, (4) structural
issues affecting many items, (5) lower-signal findings.

### 9. Synthesize and report
Generate a consolidated output covering overall status, priority items, recommended
actions, and structural patterns.

### 10. Verify completeness
Confirm all items were reviewed or marked unreviewed, all findings have evidence,
recommendations don't exceed the evidence, and the output reflects the full finding
set.

---

## Sub-agent coordination

**Parent coordinator:** defines scope and criteria, divides work, reconciles
findings, resolves or flags conflicts, produces the final output.

**Sub-agents:** review assigned items, apply criteria, return structured findings
with evidence, confidence, and recommended action.

---

## Output format

Include at minimum: Summary, Priority Queue, Summary by Criterion, Structural
Observations, and Recommended Next Actions.

---

## Guardrails

Surface items for human review — do not make final determinations. Support all
findings with evidence. State clearly when evidence is insufficient. Success is
measured by quality, traceability, and actionability — not finding count.

---

## Loaded configurations

| Operation | Reference file |
|---|---|
| Knowledge health assessment | references/knowledge-health-assessment.md |