seo-monitor

seo-monitor

Detect SEO/GEO regressions over time by snapshotting and diffing. Pure analysis — no network (the agent re-crawls; this compares).

MONITOR completion invariant: when an existing baseline is supplied, the run is incomplete until it also reads the prior ACTION-PLAN and reports every TODO as done, open, or new by comparing its Leading indicator with the fresh evidence. A baseline-only drift table is not a complete monitor delivery.

When to use

• Scheduled monitoring (e.g. an auto-task dispatches the agent in monitor mode daily/weekly).
• "Did anything regress since the baseline?" after edits or a deploy.

When NOT to use

• The first diagnosis (no baseline yet — run a full diagnose, then snapshot to set the baseline).
• Fetching pages — that is seo-crawl.

Preconditions

• snapshot: a seo-crawl JSON (optionally pass the run's --health/--geo scores to record them). compare: two snapshot files. Python 3.9+ stdlib only.

How to call

Set/refresh a baseline (store it in the project, e.g. baseline.json):

"$ORKAS_NODE" "$ORKAS_PC_DIR/bin/run-skill.cjs" seo-monitor monitor -- --op snapshot --input <crawl.json> --health <N> --geo <N> --out baseline.json

Compare a fresh snapshot against the baseline:

"$ORKAS_NODE" "$ORKAS_PC_DIR/bin/run-skill.cjs" seo-monitor monitor -- --op compare --baseline baseline.json --current current.json

Expected output

snapshot: { ok, data: { url, fetched_at, title, canonical, noindex, structured_data_types, word_count, content_hash, health_score, geo_score, ... } }.

compare:

{ "ok": true, "data": {
  "changed": true,
  "drift_findings": [ { "id": "noindex_added", "dimension": "drift", "severity": "critical",
                        "evidence": "before=False → after=True", "recommendation": "...",
                        "leading_indicator": "...", "failure_criterion": "...", "data_tier": "Measured" } ],
  "summary": { "critical": 1, "total": 1 }, "baseline_at": "...", "current_at": "...", "url": "..." } }