// Query logs from Axiom for debugging (read-only, no ingestion allowed)
| name | query-axiom-logs |
| description | Query logs from Axiom for debugging (read-only, no ingestion allowed) |
| context | fork |
| agent | Explore |
You are a log analysis specialist for the vm0 project. Your role is to query and analyze logs from Axiom for debugging purposes.
IMPORTANT: This skill is READ-ONLY. Never ingest or write data to Axiom.
Execute the following request: $ARGUMENTS
Query logs and telemetry data from Axiom using the guidelines and examples below.
There are two separate tokens for dev and prod environments:
The production token is available as the AXIOM_TOKEN shell environment variable (set by the VM environment, not in any dotfile).
# Verify it exists
echo "$AXIOM_TOKEN" | head -c 10
Dev tokens are stored in turbo/apps/web/.env.local:
AXIOM_TOKEN_SESSIONS=xaat-xxxxx
AXIOM_TOKEN_TELEMETRY=xaat-xxxxx
Ask the user to sync environment variables from 1Password:
./scripts/sync-env.sh
| Dataset | Dev Name | Prod Name | Purpose |
|---|---|---|---|
| Web Logs | vm0-web-logs-dev | vm0-web-logs-prod | Server logs (errors, warnings, API calls) |
| Agent Run Events | vm0-agent-run-events-dev | vm0-agent-run-events-prod | Agent execution events and activity |
| Sandbox System | vm0-sandbox-telemetry-system-dev | vm0-sandbox-telemetry-system-prod | Sandbox console/system logs |
| Sandbox Metrics | vm0-sandbox-telemetry-metrics-dev | vm0-sandbox-telemetry-metrics-prod | CPU, memory, disk usage |
| Sandbox Network | vm0-sandbox-telemetry-network-dev | vm0-sandbox-telemetry-network-prod | HTTP requests from sandbox |
For production datasets (-prod): use $AXIOM_TOKEN directly from the shell environment:
axiom query "APL_QUERY" -T "$AXIOM_TOKEN" -f table
For dev datasets (-dev): source the dev token first:
source turbo/apps/web/.env.local && axiom query "APL_QUERY" -T "$AXIOM_TOKEN_SESSIONS" -f table
Options:
-f table - Human-readable table (default)-f json - JSON output for processing--start-time "-1h" - Filter by time range['dataset-name']
| where condition
| project field1, field2
| limit 100
| Operator | Example |
|---|---|
| Filter | where level == "error" |
| Search | search "connection refused" |
| Time | where _time > now(-1h) |
| Select | project _time, message |
| Sort | sort by _time desc |
| Limit | limit 100 |
| Count | summarize count() by field |
axiom query "['vm0-web-logs-prod'] | where _time > now(-1h) | where level == 'error' | project _time, message, fields.context | sort by _time desc | limit 50" -T "$AXIOM_TOKEN"
axiom query "['vm0-web-logs-prod'] | search 'connection refused' | project _time, message | limit 20" -T "$AXIOM_TOKEN" --start-time "-24h"
axiom query "['vm0-agent-run-events-prod'] | where _time > now(-1h) | where eventType == 'system' | where eventData.subtype == 'error' | project _time, runId, eventData.message | limit 20" -T "$AXIOM_TOKEN"
source turbo/apps/web/.env.local && axiom query "['vm0-web-logs-dev'] | where _time > now(-1h) | where level == 'error' | project _time, message, fields.context | sort by _time desc | limit 50" -T "$AXIOM_TOKEN_SESSIONS"
source turbo/apps/web/.env.local && axiom query "['vm0-agent-run-events-dev'] | where runId == 'UUID_HERE' | sort by sequenceNumber asc" -T "$AXIOM_TOKEN_SESSIONS"
source turbo/apps/web/.env.local && axiom query "['vm0-sandbox-telemetry-system-dev'] | where runId == 'UUID_HERE' | sort by _time asc" -T "$AXIOM_TOKEN_TELEMETRY"
source turbo/apps/web/.env.local && axiom query "['vm0-sandbox-telemetry-metrics-dev'] | where runId == 'UUID_HERE' | project _time, cpu, mem_used, disk_used | sort by _time asc" -T "$AXIOM_TOKEN_TELEMETRY"
source turbo/apps/web/.env.local && axiom query "['vm0-sandbox-telemetry-network-dev'] | where _time > now(-1h) | where status >= 400 | project _time, method, url, status, latency_ms | limit 50" -T "$AXIOM_TOKEN_TELEMETRY"
| Field | Description |
|---|---|
_time | Event timestamp |
level | Log level (error, warn, info, debug) |
message | Log message |
fields.context | Context (webhook:complete, api:runs, etc.) |
vercel.environment | Vercel env (preview, production) |
vercel.region | Vercel region (iad1, etc.) |
| Field | Description |
|---|---|
_time | Event timestamp |
runId | Agent run UUID |
userId | User ID |
eventType | Type (system, assistant, tool) |
eventData.type | Subtype details |
eventData.message | Event message content |
sequenceNumber | Event sequence in run |
| Field | Description |
|---|---|
_time | Event timestamp |
runId | Agent run UUID |
userId | User ID |
log | Raw log text |
| Field | Description |
|---|---|
_time | Timestamp |
runId | Agent run UUID |
cpu | CPU usage (0-1) |
mem_total, mem_used | Memory in bytes |
disk_total, disk_used | Disk in bytes |
| Field | Description |
|---|---|
_time | Timestamp |
runId | Agent run UUID |
method | HTTP method |
url | Request URL |
status | HTTP status code |
latency_ms | Latency in milliseconds |
request_size, response_size | Bytes |
limit to avoid large result setssummarize count()) over raw queries when possibleReset turbo environment (clean node_modules, reinstall, sync DB)
Check if a PR, commit, or tag has been deployed to production
Query, investigate, and manage Sentry issues for debugging and incident response
Create or update a PR and hand it off to a coding agent worker via load balancing. Removes pending label if present, then assigns a worker.
Implement issue, review PR, and verify CI — runs issue-implement, pr-review-loop, and pr-check-loop sequentially
Deep brainstorming and solution exploration based on research findings