菜单
How to run ssmd DQ checks locally and in-cluster, interpret scores, trigger email reports, and verify results. Use when running data quality checks, re-sending DQ emails, or verifying pipeline health after deployments or backfills.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Orchestrate multi-stage pipelines for ssmd workflows. Pre-defined pipelines chain skills and agents in sequence — deploy, feature, hotfix, and dq-investigate. Use when user says "ssmdorch", "orchestrate", "pipeline", or named pipelines like "ssmdorch deploy", "ssmdorch feature", "ssmdorch hotfix", "ssmdorch dq".
Multi-agent orchestration for ssmd market data system tasks. Extends waldstorm with domain-specific experts for secmaster, data feeds, trading APIs, and data quality. Use when working on connectors, exchanges, NATS pipelines, market metadata, or user says "ssmdstorm", "market data task", "exchange work".
Multi-agent orchestration that analyzes tasks through specialized expert panels (security, performance, architecture, etc.), synthesizes prioritized actions, then plans and executes. Use when facing complex tasks requiring multiple perspectives, architectural decisions, new features needing security/performance/quality review, or user says "waldstorm", "expert review", "analyze this task".
This skill MUST be used before writing any implementation code — feature work, bug fixes, pipeline stages, data processing, API handlers, K8s manifests, or integration code. Enforces fail-loud patterns, input/output validation, connection verification, and pre-commit gates. Triggered automatically on any code writing task. Also use when user says "defensive", "fail-fast", "validate", "check failures", "harden".
Catalog of all ssmd health metrics — sources, query methods, labels, healthy/unhealthy thresholds. Covers data pipeline Prometheus metrics, HTTP endpoints, infrastructure components, DQ scores, and composite health scoring. Use when checking environment health, interpreting metrics, setting alert thresholds, or building health dashboards.
Procedures for running ssmd health checks and interpreting results. Covers quick triage, data pipeline health, infrastructure checks (NATS, Postgres, Redis), connector/archiver deep dives via port-forward, and Cloud Monitoring queries. Use when checking environment health, investigating degradation, or verifying health after deployments.
| name | ssmd-dq-run |
| description | How to run ssmd DQ checks locally and in-cluster, interpret scores, trigger email reports, and verify results. Use when running data quality checks, re-sending DQ emails, or verifying pipeline health after deployments or backfills. |
Procedures for running ssmd Data Quality checks and interpreting results.
| File | Purpose |
|---|---|
data/dq.py | DQRunner engine — 13 checks, scoring, CLI |
data/dq_email.py | Email report wrapper — runs all feeds, HTML output |
data/Dockerfile | DQ image: python:3.12-slim + duckdb + gcloud monitoring |
Requires gcloud auth application-default login for GCS access.
# Single feed
uv run data/dq.py --date 2026-02-17 --feed kalshi --stream crypto
# With verbose progress
uv run data/dq.py --date 2026-02-17 --feed kalshi --stream crypto --verbose
# JSON output (for programmatic use)
uv run data/dq.py --date 2026-02-17 --feed kalshi --stream crypto --json
# Non-default prefix (when GCS prefix differs from feed name)
uv run data/dq.py --date 2026-02-17 --feed kraken-futures --stream futures --prefix kraken-futures
uv run data/dq.py --date 2026-02-17 --feed polymarket --stream markets --prefix polymarket
Run all feeds in parallel for full pipeline verification:
uv run data/dq.py --date 2026-02-17 --feed kalshi --stream crypto
uv run data/dq.py --date 2026-02-17 --feed kraken-futures --stream futures --prefix kraken-futures
uv run data/dq.py --date 2026-02-17 --feed polymarket --stream markets --prefix polymarket
| Feed | --feed | --stream | --prefix |
|---|---|---|---|
| Kalshi | kalshi | crypto | (default: kalshi) |
| Kraken Futures | kraken-futures | futures | kraken-futures |
| Polymarket | polymarket | markets | polymarket |
The DQ CronJob runs at 03:30 UTC daily (after parquet-gen at 02:00 UTC).
Manifest: clusters/gke-prod/apps/ssmd/cronjobs/dq-daily.yaml
kubectl create job --from=cronjob/ssmd-dq-daily ssmd-dq-manual-MMDD -n ssmd
kubectl logs -n ssmd job/ssmd-dq-manual-MMDD -f
The CronJob defaults to yesterday. To override:
kubectl create job --from=cronjob/ssmd-dq-daily ssmd-dq-rerun-MMDD -n ssmd --dry-run=client -o yaml | \
sed 's|dq_email.py|dq_email.py --date 2026-02-17|' | \
kubectl apply -f -
| Grade | Score Range | Meaning |
|---|---|---|
| GREEN | >= 98 | Pipeline healthy, all checks passing |
| YELLOW | >= 85 | Minor issues, investigate when convenient |
| RED | < 85 | Significant issues, investigate promptly |
| Status | Weight | Meaning |
|---|---|---|
| pass | 1.0 | Check passed |
| warn | 0.7 | Threshold exceeded but not critical |
| fail | 0.0 | Check failed |
| skip | excluded | Not enough data to run, excluded from score |
Score = average of weights * 100.
dq.py exits 1 if any check has status faildq_email.py always exits 0 (email is the alert mechanism)from dq import DQRunner
runner = DQRunner(bucket="ssmd-data", feed="kalshi", stream="crypto")
results = runner.run("2026-02-12")
results.summary() # print human-readable report
results.score() # float 0-100
results.to_json() # JSON string
# Ad-hoc queries via the shared DuckDB connection
runner.con.execute(
"SELECT * FROM read_parquet('gcs://ssmd-data/kalshi/crypto/2026-02-12/ticker_*.parquet') LIMIT 10"
).fetchdf()
# Date range
all_results = runner.run_range("2026-02-10", "2026-02-17")
dq_email.py runs all 3 feeds, generates an HTML email with per-feed grades and check details, and sends via SMTP.
Required env vars: SMTP_USER, SMTP_PASS, SMTP_TO
Optional: SMTP_HOST (default: smtp.gmail.com), SMTP_PORT (default: 587)
These are provided in-cluster via the ssmd-smtp-credentials Secret.
After deploying a new DQ version or backfilling parquet data:
kubectl create job --from=cronjob/ssmd-dq-daily ...DQ image is built from data/Dockerfile, triggered by dq-v* tags in the 899bushwick repo (not ssmd).
See the ssmd-deploy skill for full deployment procedure.