v1.0.0 · Open profile · CC-BY-4.0

AgentMinds Reporting Profile

A thin convenience profile on top of five open standards. The wire format AgentMinds uses today, written down so external integrators don't have to reverse-engineer HTTP traffic.

Strategic position — read first

This is not a competing standard. ARP is AgentMinds' documented adoption of five existing specs, plus one extension (cross-site learned-pattern lifecycle) that no upstream standard currently covers.

If you are choosing between ARP and a real standard like OpenTelemetry GenAI or MCP, follow the standard. ARP is a profile of those standards optimized for AgentMinds' delivery surface — it is not an anchor for your own architecture.

Reorientation clause: when any upstream standard absorbs the pattern-lifecycle primitive, AgentMinds re-emits ARP within 30 days deferring to the upstream definition. We are explicitly the fastest adopter, never the spec owner.

Lineage

ARP unifies five prior-art lineages into one envelope so any agent can speak the same wire format to AgentMinds and to other observability platforms without lock-in.

Adopt: Issue lifecycle: status, first_seen, last_seen, fingerprint, level.
Why: Battle-tested across 5M+ developers.
Adopt: Metric type taxonomy (counter/gauge/histogram), UCUM units, GenAI attributes.
Why: CNCF-backed neutral telemetry. Datadog / New Relic / Honeycomb already parse it.
Adopt: Transport binding hint, tool-call shape, planned Skills-Over-MCP alignment.
Why: Anthropic-led, OpenAI / Google / Microsoft / Cursor adopters.
Adopt: SKILL.md YAML frontmatter (name, description, allowed-tools).
Why: De-facto skill discovery format used by Claude Code agents.
Adopt: Descriptor envelope (media-type, digest, size), metric shape, observability data referencing.
Why: Linux Foundation hosted. 75+ member orgs; Cisco / Dell / Google Cloud / Oracle / Red Hat.

The one extension we own

§4.1 Pattern object — cross-site learned-pattern lifecycle. AGNTCY and OASF describe agents; nobody has standardized how agents share what they learned across sites.

Every pattern carries a stable fingerprint, status (active / solved / obsolete), first_seen / last_seen, applicable_site_types, and confidence. That single primitive is what enables AgentMinds' cross-site recommendation pool — and it is the only piece of ARP that is not a verbatim adoption of the lineage.

Conformance levels

LevelRequired features
L0Envelope + report.summary + warnings[].message
L1L0 + severity + category + status per warning
L2L1 + fingerprint + first_seen + last_seen + count
L3L2 + memory.learned_patterns with §4.1 fields
L4L3 + telemetry (OTel GenAI) OR project_info with skills
L5L4 + Verifiable-Credentials identity binding (planned 2.0)

AgentMinds' collector requires L0, recommends L2; cross-site recommendation filtering unlocks at L2+.

Documentation

Read the spec

ARP is published under CC-BY-4.0. Reuse and adapt freely; attribution to AgentMinds appreciated.