feat: canonical ARD identifiers and six source-pinned OpenEnv records

[thegovind]

Summary

Accept canonical urn:air: contributor identifiers alongside legacy urn:ai:,
normalize once, and detect collisions after normalization. Existing legacy
inputs remain valid.

Propose six source-pinned OpenEnv Environment records for task-driven discovery:

Record	What someone can find
Echo smoke test	Client and tool-call smoke testing
Python code execution	Run Python snippets and inspect execution output
BrowserGym web tasks	Browser interaction and form-filling tasks
Calendar scheduling	Simulated event creation and scheduling
Chess environment	Legal UCI moves against an opponent
Reasoning Gym questions	Dataset-scored, single-question episodes

All six come from one complete 39-card producer snapshot at proposed OpenEnv
source revision 338577885c00d69a92e652991b367fa246918093. Each record preserves
the original complete data, GitHub source URI, environment path and full SHA.
Friendly display names and an rl-environment tag aid browsing.

Snapshot digest:
sha256:105b0d254b72b599f1544b371d50fb7e0f0c3f7badf224b91791db7868ee929a.

The previous Echo-only proposal is replaced by the same-current-snapshot Echo
record. That older proposal was never admitted, so this is not withdrawal of an
already published listing.

Scope and authority

• Generic canonical/legacy identifier compatibility and duplicate detection.
• Contributor guidance for complete inline extension records.
• Six application/vnd.openenv.environment-card+json examples.
• Canonically regenerated ai-catalog.json, including the existing public MCP
  supplement. The generated file is not hand-edited.

The GitHub-hosted publisher namespace is proposed for catalog-owner review,
not asserted as a verified endorsement. Source ownership and publication
authority remain distinct. Source-code license evidence does not license
external datasets or dependencies.

These records describe environments. They are not MCP installation endpoints,
training actions, runtime validation reports or execution approvals. Calendar
advertises only five source-evidenced event tools. No interfaces are invented
for the other newly authored task families.

Validation

• All 27 catalog tests pass, including canonical/legacy aliases, normalized
  duplicate identities, inline data preservation, distinct revisions, contributor
  templates and the six-record single-snapshot invariant.
• Generation and --check agree: 2,113 entries.
• All six actual payloads pass OpenEnv's packaged JSON schemas and relational
  source/path/revision validation, plus the independent HF consumer's parser.
• Every complete card in the 39-card snapshot round-trips unchanged through
  six HF HTTP pages and six MCP pages.
• Five held-out, non-Echo task queries select their intended environments through
  the OpenEnv baseline, HF HTTP and HF CLI. They are not copies of the authored
  representative queries and are not a ranking-quality benchmark.
• Whitespace checks pass. Current upstream catalog changes were merged, and
  generated-file conflicts were resolved with the existing generator.

Dependencies

The design is in merged huggingface/OpenEnv#1093. The producer and task metadata
are proposed in huggingface/OpenEnv#1138; the independent consumer is
huggingface/hf-discover#46.

Merged ards-project/ard-spec#85 improves diagnostics. It is complementary, not an
admission prerequisite. This catalog PR still requires owner admission. A merge
does not itself establish downstream ingestion, first-class presentation,
deployment, or live Agent Finder availability.

[Copilot]

Copilot review overview

🟡 Changes recommended

Align the contributor template and field table with canonical urn:air: and type fields.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review tier: Lite
Findings: 1

New issues introduced by this change (1)

Severity	Finding
	CONTRIBUTING.md — This new guidance is contradicted by the contributor template and field table below, which still…

What changed in this PR

Adds canonical ARD identifier support with legacy compatibility and a pinned OpenEnv Echo catalog entry.

Changes:

• Normalizes identifiers and detects duplicate identities.
• Adds regression tests and contributor guidance.
• Adds the Echo record and regenerates the catalog.

File	Summary
tests/​test_extension_catalog.py	Tests identifier compatibility and revision handling.
scripts/​generate_ai_catalog.py	Adds normalization and collision handling.
CONTRIBUTING.md	Updates contributor guidance.
catalog/​huggingface/​openenv-echo-8e6e2770.json	Adds the pinned Echo environment card.
ai-catalog.json	Regenerated catalog output.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[thegovind]

Fixed in 285cd4c. The template and field table now use canonical urn:air/type, show owner/repository source metadata, and distinguish URL delivery from inline data. A regression test parses and validates the actual documented template. This also exposed a canvas-only validator that still required the legacy field; it now accepts canonical type while rejecting conflicting aliases. All 23 generator tests pass.