NATS Management Skill

Analyze and manage NATS clusters with safe, read-only operations.

MANDATORY: Two-Phase Execution

You MUST follow this two-phase pattern. Skipping Phase 1 causes hallucinated stream/consumer names.

Phase 1: Discovery (ALWAYS run first)

#!/bin/bash

# 1. Server info
nats server info --server "$NATS_URL"

# 2. Account info
nats account info --server "$NATS_URL"

# 3. List JetStream streams
nats stream list --server "$NATS_URL"

# 4. List consumers for a stream
nats consumer list MY_STREAM --server "$NATS_URL"

# 5. Stream info
nats stream info MY_STREAM --server "$NATS_URL"

Phase 1 outputs:

• Server version and cluster info
• JetStream streams and consumers
• Stream subjects and message counts

Phase 2: Analysis (only after Phase 1)

Only reference streams, consumers, and subjects confirmed in Phase 1.

Shell Script Patterns

Helper Function

#!/bin/bash

# Core NATS CLI helper — always use this
nats_cmd() {
    nats --server "${NATS_URL:-nats://localhost:4222}" "$@"
}

# NATS monitoring API
nats_monitor() {
    local endpoint="$1"
    curl -s "http://${NATS_MONITOR_HOST:-localhost}:8222/$endpoint"
}

Anti-Hallucination Rules

• NEVER reference a stream without confirming via nats stream list
• NEVER reference a consumer without confirming via nats consumer list
• NEVER assume subject names — always check stream configuration
• NEVER guess cluster size — always check server info
• NEVER assume JetStream is enabled — verify with account info

Safety Rules

• READ-ONLY ONLY: Use only info, list, report, server commands, monitoring API GET
• FORBIDDEN: stream add/edit/delete, consumer add/delete, pub without explicit user request
• NEVER subscribe to production subjects for analysis — use monitoring endpoints
• Use nats server report for cluster-wide analysis

Common Operations

Cluster Health Overview

#!/bin/bash
echo "=== Server Info ==="
nats_cmd server info --json | jq '{server_id, version, go, host, port, max_payload, proto, jetstream, cluster: .connect_urls}'

echo ""
echo "=== Server List ==="
nats_cmd server list

echo ""
echo "=== Account Info ==="
nats_cmd account info --json | jq '{memory, storage, streams, consumers, api: {total, errors}}'

echo ""
echo "=== Connection Report ==="
nats_cmd server report connections

JetStream Stream Analysis

#!/bin/bash
echo "=== Streams ==="
nats_cmd stream list --json | jq '.[] | {name: .config.name, subjects: .config.subjects, messages: .state.messages, bytes: .state.bytes, consumers: .state.consumer_count, retention: .config.retention, storage: .config.storage}'

echo ""
echo "=== Stream Details ==="
STREAM="${1:-MY_STREAM}"
nats_cmd stream info "$STREAM" --json | jq '{config: {name: .config.name, subjects: .config.subjects, retention: .config.retention, max_msgs: .config.max_msgs, max_bytes: .config.max_bytes, max_age: .config.max_age, replicas: .config.num_replicas}, state: .state}'

echo ""
echo "=== Stream Report ==="
nats_cmd stream report

Consumer Status

#!/bin/bash
STREAM="${1:-MY_STREAM}"

echo "=== Consumers for $STREAM ==="
nats_cmd consumer list "$STREAM" --json | jq '.[] | {name: .config.name, durable_name: .config.durable_name, deliver_policy: .config.deliver_policy, ack_policy: .config.ack_policy, num_pending: .num_pending, num_redelivered: .num_redelivered, num_ack_pending: .num_ack_pending}'

echo ""
echo "=== Consumer Details ==="
CONSUMER="${2:-MY_CONSUMER}"
nats_cmd consumer info "$STREAM" "$CONSUMER" --json | jq '{config, delivered: .delivered, ack_floor: .ack_floor, num_pending, num_redelivered, num_ack_pending}'

echo ""
echo "=== Consumer Report ==="
nats_cmd consumer report "$STREAM"

Monitoring Endpoints

#!/bin/bash
echo "=== Varz (general stats) ==="
nats_monitor "varz" | jq '{server_id, version, uptime, mem, cpu, connections, total_connections, subscriptions, slow_consumers, in_msgs, out_msgs, in_bytes, out_bytes}'

echo ""
echo "=== Connz (connections) ==="
nats_monitor "connz?sort=msgs_to&limit=10" | jq '.connections[] | {cid, name, ip, subscriptions, msgs_to, msgs_from, bytes_to, bytes_from, uptime}'

echo ""
echo "=== Routez (cluster routes) ==="
nats_monitor "routez" | jq '.routes[]? | {rid, remote_id, ip, port, in_msgs, out_msgs}' 2>/dev/null

echo ""
echo "=== JetStream Info ==="
nats_monitor "jsz" | jq '{server_id, config, streams, consumers, messages, bytes, api: {total, errors}}'

Output Format

Present results as a structured report:

Managing Nats Report
════════════════════
Resources discovered: [count]

Resource       Status    Key Metric    Issues
──────────────────────────────────────────────
[name]         [ok/warn] [value]       [findings]

Summary: [total] resources | [ok] healthy | [warn] warnings | [crit] critical
Action Items: [list of prioritized findings]

Target ≤50 lines of output. Use tables for multi-resource comparisons.

Counter-Rationalizations

Shortcut	Counter	Why
"I'll skip discovery and check known resources"	Always run Phase 1 discovery first	Resource names change, new resources appear — assumed names cause errors
"The user only asked for a quick check"	Follow the full discovery → analysis flow	Quick checks miss critical issues; structured analysis catches silent failures
"Default configuration is probably fine"	Audit configuration explicitly	Defaults often leave logging, security, and optimization features disabled
"Metrics aren't needed for this"	Always check relevant metrics when available	API/CLI responses show current state; metrics reveal trends and intermittent issues
"I don't have access to that"	Try the command and report the actual error	Assumed permission failures prevent useful investigation; actual errors are informative

Common Pitfalls

• Core NATS vs JetStream: Core NATS is fire-and-forget; JetStream provides persistence — know which you need
• Consumer ack pending: High ack_pending means consumers are slow — check max_ack_pending limits
• Stream retention: Limits (max_msgs, max_bytes, max_age) silently discard old messages
• Slow consumers: NATS disconnects slow consumers — monitor slow_consumers metric
• Subject wildcards: > matches all remaining tokens; * matches one token — test patterns carefully
• Redelivery loops: Messages failing ack cause redelivery — set max_deliver to prevent infinite loops