+++ name = "quality-review" description = "Review merge quality and track per-worker trends" version = 1

[gate] type = "cooldown" duration = "6h"

[tracking] labels = ["plugin:quality-review", "category:quality"] digest = true

[execution] timeout = "5m" notify_on_failure = true severity = "medium" +++

Quality Review — Trend Analysis

This plugin runs every 6h during Deacon patrol. It analyzes quality-review result wisps recorded by the Refinery during merges, computes per-worker trends, and alerts on quality breaches.

Step 1: Query recent quality-review results

Fetch all quality-review result wisps from the last 24 hours:

bash

bd list --json --all -l type:plugin-run,plugin:quality-review-result --created-after=-24h

If no results are found, record a run wisp and stop:

bash

bd create "quality-review: No results in last 24h" -t chore --ephemeral \
  -l type:plugin-run,plugin:quality-review,result:success \
  -d "No quality-review results in last 24h. Nothing to analyze." \
  --silent 2>/dev/null || true

Step 2: Compute per-worker trends

Parse the wisp labels to extract per-worker data. Each result wisp has labels:

worker:<polecat-name>
rig:<rig-name>
score:<0.0-1.0>
recommendation:<approve|request_changes>

For each worker, compute:

Average score across all results in window
Rejection rate: count of recommendation:request_changes / total
Trend direction: Compare first-half avg vs second-half avg of the window
- Difference > 0.05: improving
- Difference < -0.05: declining
- Otherwise: stable

Step 3: Classify worker status

Apply thresholds to each worker's average score:

OK: avg >= 0.60
WARN: 0.45 <= avg < 0.60
BREACH: avg < 0.45

Step 4: Alert on breaches

For each worker in BREACH status, send an alert:

bash

gt mail send mayor/ -s "Quality BREACH: <worker>" -m "Worker: <worker>
Rig: <rig>
Avg Score: <avg>
Reviews: <count>
Rejection Rate: <rate>%
Trend: <improving|stable|declining>

Action: Review recent merges from this worker for quality issues."

Also escalate:

bash

gt escalate "Quality BREACH: <worker> (avg: <avg>)" \
  --severity medium \
  --reason "Worker <worker> in rig <rig> has avg quality score <avg> over <count> reviews"

Step 5: Record run result

Record a summary wisp for this plugin run:

bash

bd create "quality-review: Analyzed <N> workers over <M> reviews" -t chore --ephemeral \
  -l type:plugin-run,plugin:quality-review,result:success \
  -d "Analyzed <N> workers over <M> reviews. <B> breaches, <W> warnings." \
  --silent 2>/dev/null || true

If any step fails unexpectedly, record a failure wisp and escalate:

bash

bd create "quality-review: FAILED" -t chore --ephemeral \
  -l type:plugin-run,plugin:quality-review,result:failure \
  -d "<error description>" \
  --silent 2>/dev/null || true

gt escalate "Plugin FAILED: quality-review" \
  --severity medium \
  --reason "$ERROR"

How scores get recorded (reference)

This plugin does NOT record scores itself. The Refinery records result wisps during merges via the quality-review formula step. Each merge produces a wisp like:

bash

bd create "quality-review: Score 0.85, approve" -t chore --ephemeral \
  -l type:plugin-run,plugin:quality-review-result,worker:<polecat-name>,rig:<rig-name>,score:0.85,recommendation:approve,result:success \
  -d "Score: 0.85, approve. Issues: 1 minor (style)" \
  --silent 2>/dev/null || true

This creates the data that Step 1 queries.