[dmealing]

What Changed

• Re-checks the fit-assessment capability sheet (agent-context/skills/metaobjects-fit-assessment/references/capability-sheet.md) against 1.2.0 / Maven 8.2.0. Rewrites the metadata-sharing bullet: TypeScript publishes; TypeScript, Python, Java, Kotlin and C# consume a build-time snapshot. Maven and NuGet transports are design only. Also records 1.2 facts: bare verify runs codegen, template and requirement gates, Postgres shadow-DB drift, report-export CSV/JSON/XML, credential and @formExclude handling, C# namespace mapping, and Jinja templates for Python and TypeScript only.
• Updates the SKILL.md capability-sheet heading and "last checked" line to 1.2.0 / Maven 8.2.0.
• Mirrors the sharing-model correction in README.md (dependencies row), docs/llms/llms.txt and docs/llms/llms-full.txt.

Validation

Risk Assessment

✅ Low: Doc-only restamp. I checked each new capability claim against docs/features (metadata-dependencies, cli, api-contract, reporting, migrations-and-drift), docs/ports/csharp.md, the C# TemplateEngines source, and CONFORMANCE.md, and found no mismatch. No stale 1.1.4 or 8.1.4 strings remain under agent-context, and no checked-in fixture depends on the sheet.

Testing

Docs-only change (agent-context skill markdown). One live check: the assess-prompt builder (node scripts/test-build-assess-prompt.mjs) run against the real SKILL.md and capability-sheet.md, all checks pass. Unit/conformance suites (bun test in server/typescript/packages/sdk, 92 pass) and a static grep over fixtures/ and sdk were run but are not live product exercise. Per-port 1.2.0 claims were not driven live; that needs a Maven/dotnet build of released ports, not run here. Verdict inconclusive: one live scenario passed, cross-port scenarios untested.

• Live validation: ⚠️ inconclusive - 1 of 4 scenarios driven live against the product

Scenario	Result	Live	Evidence
Assess prompt builder inlines the updated capability sheet under its new 1.2.0 heading	✅ pass	live	node scripts/test-build-assess-prompt.mjs — 'the capability sheet is inlined under its heading' and 'hosted form is the whole prompt' ok
Agent-context conformance and grounding suites accept the re-checked sheet and SKILL.md	⏸️ untested	no	Prior payload recorded bun unit/conformance runs (92 pass) only; those are not a live run of the product, so no live result is established.
Cross-port 1.2.0 claims in the sheet (Java/Kotlin/C# dependency consumption, C# namespace from packages, report export formats) match the released ports	⏸️ untested	no	Verifying needs a Maven/dotnet build of the released 1.2.0 ports; not run. Claims were traced to the docs the sheet cites, not re-driven.
No conformance fixture depends on the fit-assessment capability sheet, so none needed regeneration	⏸️ untested	no	Prior payload recorded only a static grep over fixtures/ and sdk; that is a dependency lookup, not a live run, so no live result is established.

• Outcome: ⏭️ skipped across 1 run (15m46s)

Pipeline

Updates from git push no-mistakes

✅ **Intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

⏭️ **Test** - skipped

• ⚠️ agent-context/skills/metaobjects-fit-assessment/references/capability-sheet.md:26 - Sheet claims for 1.2.0 (dependency consumption in Java/Kotlin/C#, C# namespace derivation, Jinja engine split, report CSV/JSON/XML export, shadow-DB drift) were cross-checked against the docs the sheet cites, not re-derived from 1.2.0 source or run against the released ports. Live check of cross-port claims was not done.
• ⚠️ live validation verdict: inconclusive (1 of 4 scenarios were driven live against the product); untested: Agent-context conformance and grounding suites accept the re-checked sheet and SKILL.md, Cross-port 1.2.0 claims in the sheet (Java/Kotlin/C# dependency consumption, C# namespace from packages, report export formats) match the released ports, No conformance fixture depends on the fit-assessment capability sheet, so none needed regeneration
• Live validation: ⚠️ inconclusive - 1 of 4 scenarios driven live against the product

Scenario	Result	Live	Evidence
Assess prompt builder inlines the updated capability sheet under its new 1.2.0 heading	✅ pass	live	node scripts/test-build-assess-prompt.mjs — 'the capability sheet is inlined under its heading' and 'hosted form is the whole prompt' ok
Agent-context conformance and grounding suites accept the re-checked sheet and SKILL.md	⏸️ untested	no	Prior payload recorded bun unit/conformance runs (92 pass) only; those are not a live run of the product, so no live result is established.
Cross-port 1.2.0 claims in the sheet (Java/Kotlin/C# dependency consumption, C# namespace from packages, report export formats) match the released ports	⏸️ untested	no	Verifying needs a Maven/dotnet build of the released 1.2.0 ports; not run. Claims were traced to the docs the sheet cites, not re-driven.
No conformance fixture depends on the fit-assessment capability sheet, so none needed regeneration	⏸️ untested	no	Prior payload recorded only a static grep over fixtures/ and sdk; that is a dependency lookup, not a live run, so no live result is established.

• scripts/ci-local.sh --only ts-fast --only ts-unit --strict-toolchains
• node scripts/test-build-assess-prompt.mjs (real prompt builder over SKILL.md + capability-sheet.md; all checks pass)
• bun test test/agent-context-conformance.test.ts test/agent-context-capability-grounding.test.ts test/agent-context test/agent-docs-subpath-retired.test.ts (in server/typescript/packages/sdk; 92 pass)
• git status clean after runs; no fixture references fit-assessment or capability-sheet

⚠️ **Document** - 2 warnings

• ⚠️ AGENTS.md:42 - AGENTS.md 'Where the versions are' still says latest is 1.1.4 / 8.1.4 and omits 1.2.0. Release-wide restamp, not this sheet change; AGENTS.md is a human-owned agent memory file.
• ⚠️ README.md:163 - README implementation table still pins 1.1.4 / 8.1.4 per registry (npm, Maven, NuGet, PyPI). Stale after the 1.2.0 release, outside the sheet change.
• ⚠️ docs/llms/llms.txt:66 - llms.txt 'Sharing a model' says Java, Kotlin and C# arrive in Phase 2. Stale since 1.2.0 shipped consumption; the mirror is copied by scripts/finish-release.mjs, so it may be restamped by the release process. llms-full.txt line 120 has the same claim.

🔧 Fix applied.
2 warnings still open:

• ⚠️ AGENTS.md:42 - 'Where the versions are' still says latest is 1.1.4 / 8.1.4 and omits 1.2.0; human-owned agent memory file, release-wide restamp.
• ⚠️ README.md:163 - Implementation table still pins 1.1.4 / 8.1.4 per registry; stale after 1.2.0, outside the sheet change.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

Follow-up (separate step)

A new skill tag plus a site link update (the metaobjects.dev link still points at the skill's v1.1.4 tag) is out of scope here.