ci: sync sample catalog from foundry-samples (yimin/catalog-review-correction-e2e-20260928)

[foundry-toolkit-sync-bot]

Summary

This automated draft PR incrementally updates samples/hosted-agent/sample-catalog.json
from a pinned scan of microsoft-foundry/foundry-samples.
The single snapshot keeps template facts in templates and grouping, Patterns,
and Details in cards/patterns. Clients read one Raw file without a REST revision lookup.

What changed

• Surviving template metadata, card IDs, titles, Patterns, and relative ordering are preserved.
• Cards with unchanged membership keep their Details verbatim and require no review call.
• Deleted templates and their card references are removed together; empty cards and unused dimension options are removed.
• New template metadata is generated by the existing AI service from its README; dimensions come from source metadata.
• AI prefers existing cards with the same core task, title and Pattern. Occupied language/framework/protocol tuples cannot be reused within a card.
• Proposed new cards are independently compared with every tuple-compatible candidate in the same Pattern, including cards created earlier in the run. A same-task result reuses the candidate; SDK or protocol differences alone do not justify a new card.
• After grouping, membership-changed cards receive an explicit keep/change audit of all eight Details fields against the final members' pinned READMEs. Audit actions must match the sparse patch. Both reviews select supplied README excerpt IDs; code resolves them to original text instead of asking AI to copy quotes. Reviews and changed fields are logged.
• Valid Requirements are preserved unless factually inaccurate. Commas do not create extra values, and requiresModel=false does not imply model access is unnecessary. Variant-specific approval/recovery claims must be qualified; generated output describes one selected implementation.
• Deletion-only membership changes also require Details review for surviving cards; removing a whole card does not.
• New cards are created only when no compatible candidate fits; their Details and one primary Pattern are AI-generated for review.
• The catalog commitSha advances only when templates are added or removed. It also pins the implementations of surviving templates to that revision.
• Failed AI generation/review, incomplete audits, unknown README references, incomplete scans, or invalid grouping stop the job before PR creation. Verified source references do not replace human review of semantic accuracy.
• A final bounded agent step reads the repository review skill and fixed-source evidence, then proposes prose corrections to this Draft PR. Trusted code validates scope, evidence references and tests before appending a commit. Failed or unresolved review leaves the Draft PR open and the workflow blocked; no approval or merge is automated.

Reviewer Checks

• Confirm the catalog is self-contained and card-only edits do not change template facts.
• Confirm new templates have correct language, framework, and protocol values.
• Review the auto-generated description for any new templates.
• Confirm no PM-edited display fields were accidentally lost.
• Confirm every template belongs to one card and each card's selection combinations are unique.
• Review card Details and Pattern assignments when their template membership changes.
• Confirm existing-card reuse preserves the core task and Details edits are minimal and accurate for every variant; uniqueness is enforced by code, not left to AI.

[foundry-toolkit-sync-bot]

Automated Catalog Review: passed

Input: 8572518

Output: 9045400

Skill SHA-256: 103ad8f96cc30df1b99754f75c88b733908235cf3c647274006d8c228f4664f6

Phase: complete

Completed passes: 2

Proposed corrections passed scope checks, structural validation and regression tests.

The PR remains draft. This is automated evidence-assisted review, not human approval or runtime deployment validation.

[Yimin-Jin]

Closing this isolated correction-path fixture without merging. The final test passed at https://github.com/microsoft/foundry-dev-tools/actions/runs/36383059780 with automated correction commit 9045400: description 193 -> 81 characters, all other catalog values unchanged; 137 regression tests passed. The implementation remains in #825. This verifies execution and protected edit scope, not human approval of AI prose; the generated description still repeats language/protocol wording. Test branches are being removed; this PR and CI report retain the evidence.