Estuary task stats

Skill estuary/agent-skills/skills/estuary-task-stats

Agent skills for setting up and operating Estuary data pipelines through your AI assistant.

Install

npx -y skills add estuary/agent-skills --skill estuary-task-stats

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Check task processing statistics to see how much data and how many docs are moving. Use when checking throughput, diagnosing a stalled task, verifying data movement, or reporting on task activity. Use when user says "is data flowing", "how much data", "check throughput", "how many documents", "task not processing", "data not arriving", "stats", "bytes processed", "stalled capture", "no data in destination", "data volume", "task report", "usage report", or "how much data did we process".

SKILL.md

5.3 KB, as published. Nobody here has run it

flowctl raw stats - Task Processing Statistics

Concepts: Stats show document and byte counts per transaction for a task. They answer "is data actually moving, and how much?" — separate from flowctl catalog status, which only shows control-plane state. A task can show OK status while having zero recent stats (stalled, no new source data, or sync schedule delay).

IMPORTANT: The command is flowctl raw stats, not flowctl stats.

Prerequisites

flowctl must already be authenticated — see the estuary-flowctl-setup skill.

Total Data Over a Period

Materialization — bytes read from source collections ("Data Read" in UI):

flowctl raw stats --task <task> --since 24h | \
  jq -s '{
    total_gb:     ([.[] | select(.materialize != null) | .materialize | to_entries[] | .value.right.bytesTotal // 0] | add // 0 | . / 1073741824 * 10 | round / 10),
    total_docs:   ([.[] | select(.materialize != null) | .materialize | to_entries[] | .value.right.docsTotal  // 0] | add),
    transactions: ([.[] | select(.materialize != null)] | length)
  }'

Capture — bytes written to Estuary collections ("Data Written" in UI):

flowctl raw stats --task <task> --since 24h | \
  jq -s '{
    total_gb:     ([.[] | select(.capture != null) | .capture | to_entries[] | .value.out.bytesTotal // 0] | add // 0 | . / 1073741824 * 10 | round / 10),
    total_docs:   ([.[] | select(.capture != null) | .capture | to_entries[] | .value.out.docsTotal  // 0] | add),
    transactions: ([.[] | select(.capture != null)] | length)
  }'

Hourly Breakdown

Useful for spotting gaps, spikes, or batch patterns. Matches the hourly bar chart in the Estuary UI.

Materialization:

flowctl raw stats --task <task> --since 48h | \
  jq -s '[.[] | select(.materialize != null) | {
    hour:  .ts[0:13],
    bytes: ([.materialize | to_entries[] | .value.right.bytesTotal // 0] | add),
    docs:  ([.materialize | to_entries[] | .value.right.docsTotal  // 0] | add)
  }] | group_by(.hour) | map({
    hour: .[0].hour,
    gb:   (([.[].bytes] | add) / 1073741824 * 10 | round / 10),
    docs: ([.[].docs]  | add)
  })'

Capture:

flowctl raw stats --task <task> --since 48h | \
  jq -s '[.[] | select(.capture != null) | {
    hour:  .ts[0:13],
    bytes: ([.capture | to_entries[] | .value.out.bytesTotal // 0] | add),
    docs:  ([.capture | to_entries[] | .value.out.docsTotal  // 0] | add)
  }] | group_by(.hour) | map({
    hour: .[0].hour,
    gb:   (([.[].bytes] | add) / 1073741824 * 10 | round / 10),
    docs: ([.[].docs]  | add)
  })'

Hours with no entry = no transactions that hour (expected for gaps, quiet sources, or batch sync schedules).

Stats Structure

Raw entries are nested by task type, then by collection/binding:

{
  "ts": "2026-04-10T00:01:00Z",
  "capture":     { "<collection>": { "out":   { "docsTotal": N, "bytesTotal": N } } },
  "materialize": { "<collection>": { "right": { "docsTotal": N, "bytesTotal": N },
                                     "left":  { "docsTotal": N, "bytesTotal": N },
                                     "out":   { "docsTotal": N, "bytesTotal": N } } },
  "derivation":  { "<transform>":  { "input": { "docsTotal": N }, "out": { "docsTotal": N } } }
}

For materializations:

right — docs/bytes read from source Estuary collections ("Data Read" in UI)
left — docs/bytes read from the destination (for merging)
out — docs/bytes written to the destination

The raw output also contains interval heartbeat entries and ack entries — the select(.materialize != null) / select(.capture != null) filters exclude these.

Diagnosing a Stalled Task

If flowctl catalog status shows OK but data isn't arriving:

# Check for any recent activity
flowctl raw stats --task <task> --since 6h | jq -s '[.[] | select(.materialize != null)] | length'
# → 0 = no transactions in 6h despite OK status

# Run hourly breakdown to find where the gap starts (substitute .capture for capture tasks)
flowctl raw stats --task <task> --since 48h | \
  jq -s '[.[] | select(.materialize != null) | {hour: .ts[0:13], gb: ([.materialize | to_entries[] | .value.right.bytesTotal // 0] | add // 0 | . / 1073741824 * 10 | round / 10)}] | group_by(.hour) | map({hour: .[0].hour, gb: ([.[].gb] | add)})'

Common causes of stats silence with OK status:

Source has no new data (expected for quiet sources)
Sync schedule delay — materializations batch on a schedule (default: 30 min)
Connector in retry backoff — check with flowctl logs --task <task> --since 2h

Related Skills

estuary-task-health — Single-task health check combining status, stats, logs, and history
estuary-catalog-status — Check control-plane state (running/failed/disabled)
estuary-logs — Investigate errors if stats show zero activity