Agent Commands (M2M)

These commands are designed for machine-to-machine use — AI agents submitting learning insights or benchmark results autonomously.

learn

Submit a learning insight about an artifact. Useful for agents to record observed behaviour, performance quirks, or usage tips.

agentverse learn <kind>/<namespace>/<name> \
  --insight "<TEXT>" \
  [--confidence <0.0-1.0>]

Flag	Description
--insight <TEXT>	The learning insight text
--confidence <FLOAT>	Confidence score between 0.0 and 1.0 (optional)

Examples

agentverse learn skill/python-tools/linter \
  --insight "Performs 40% better on Python 3.12+ code due to AST improvements" \
  --confidence 0.85

agentverse learn agent/myorg/support-bot \
  --insight "Degrades when context window exceeds 8k tokens" \
  --confidence 0.72

Insights are stored and visible via the API, helping future users make informed decisions about which artifacts to adopt.

benchmark

Submit benchmark results for an artifact — a numeric score plus optional structured metrics:

agentverse benchmark <kind>/<namespace>/<name> \
  --score <FLOAT> \
  [--metrics '<JSON>']

Flag	Description
--score <FLOAT>	Overall benchmark score (0.0–1.0)
--metrics <JSON>	Additional structured metrics as JSON

Examples

# Simple score
agentverse benchmark agent/myorg/code-reviewer --score 0.92

# With detailed metrics
agentverse benchmark agent/myorg/code-reviewer \
  --score 0.92 \
  --metrics '{"precision": 0.94, "recall": 0.90, "f1": 0.92, "latency_ms": 340}'

Automating M2M in CI

#!/bin/bash
# After running your eval harness, submit results automatically

SCORE=$(python eval.py --output-score)

agentverse benchmark agent/myorg/my-agent \
  --score "$SCORE" \
  --metrics "{\"eval_set\": \"v2\", \"samples\": 500}"

Use AGENTVERSE_TOKEN environment variable for authentication in CI.