Agent Metrics

Agent Metrics

Generate throughput and quality reports from the coordinator's audit trail and episodic memory. Provides three modes: throughput overview, failure rate analysis, and capability gap frequency.

Arguments

$ARGUMENTS - Optional flags:

• --time-range <days> (default: 30)
• --failures (show failure rate analysis by agent type, skill, failure_type)
• --gaps (show capability gap frequency, cross-referenced with /improve-harness reports)
• --output <path> (write report to file; default: stdout)

Modes

Default: Throughput Report

Queries the audit trail for operational metrics:

• Tasks completed / failed
• PRs opened
• Review cycles per PR
• Average time-to-merge

--failures: Failure Rate Analysis

Queries episodic memory for failure patterns and computes:

• Failure rates by failure_type (timeout, scope_violation, etc.)
• Failure rates by affected skill
• Trends (increasing/decreasing)

--gaps: Capability Gap Frequency

Queries episodic memory for capability_gap entries and:

• Ranks gaps by frequency
• Cross-references with /improve-harness reports if available

Prerequisites

• Python 3.11+
• Coordinator running at COORDINATOR_URL (default: http://localhost:8000)
• Falls back gracefully with a warning if the coordinator is unreachable

Steps

1. Query Metrics

python3 <agent-skills-dir>/agent-metrics/scripts/query_metrics.py \
  --time-range ${TIME_RANGE:-30} \
  ${FAILURES:+--failures} \
  ${GAPS:+--gaps} \
  --json

2. Generate Dashboard

python3 <agent-skills-dir>/agent-metrics/scripts/generate_dashboard.py \
  --time-range ${TIME_RANGE:-30} \
  ${FAILURES:+--failures} \
  ${GAPS:+--gaps} \
  ${OUTPUT:+--output "$OUTPUT"}