Two coupled changes:

1. Pure BOM regen of docs/user/container-images.md via `make bom-docs`
   to reflect current registry + chart state. No code changes, no
   registry pins moved; this picks up drift accumulated since main's
   last BOM commit:
   - slinky-slurm-operator and slinky-slurm-operator-crds BOM rows
     added (registry entries landed in NVIDIA#866 without a corresponding
     BOM regen)
   - kubeflow-trainer's bundled pytorch image bumped upstream
   - busybox bumped upstream in some chart's rendered templates

2. Add the BOM-regen rule to contributor and agent docs, surfaced by
   the same NVIDIA#872 review that flagged the drift:
   - .claude/CLAUDE.md (auto-synced to AGENTS.md) — short rule in the
     "Common Tasks" section: after any change to recipes/registry.yaml,
     a component values file, or a chart pin, run `make bom-docs` and
     commit the regenerated container-images.md in the same PR
   - docs/contributor/data.md — new "Regenerating the BOM" subsection
     under Component Configuration covering when to run it, how it can
     surface upstream chart drift, and how to handle unrelated drift
     (split into a catch-up PR vs land it together)

Surfaced as an unrelated diff in NVIDIA#872 (aws-efa v0.5.26 bump) when
`make bom-docs` was rerun against the rebased branch. Splitting this
catch-up out so NVIDIA#872's diff is scoped to aws-efa content.

Open question for follow-up (not blocking this PR): how did NVIDIA#866 merge
with a stale BOM? `make qualify` enforces BOM-up-to-dateness locally,
so either the CI check is path-conditional and skipped on that PR, or
there's another loophole. Worth a separate audit.

Test plan:
- `make qualify` passes (Go tests, golangci-lint + yamllint, BOM
  staleness check, agents-sync check, chart-pin verification, 20/20
  chainsaw, vulnerability scan, license headers)
- No code changes, no chart pins moved, no behavior change

Two coupled changes:

1. Pure BOM regen of docs/user/container-images.md via `make bom-docs`
   to reflect current registry + chart state. No code changes, no
   registry pins moved; this picks up drift accumulated since main's
   last BOM commit:
   - slinky-slurm-operator and slinky-slurm-operator-crds BOM rows
     added (registry entries landed in NVIDIA#866 without a corresponding
     BOM regen)
   - kubeflow-trainer's bundled pytorch image bumped upstream
   - busybox bumped upstream in some chart's rendered templates

2. Add the BOM-regen rule to contributor and agent docs, surfaced by
   the same NVIDIA#872 review that flagged the drift:
   - .claude/CLAUDE.md (auto-synced to AGENTS.md) — short rule in the
     "Common Tasks" section: after any change to recipes/registry.yaml,
     a component values file, or a chart pin, run `make bom-docs` and
     commit the regenerated container-images.md in the same PR
   - docs/contributor/data.md — new "Regenerating the BOM" subsection
     under Component Configuration covering when to run it, how it can
     surface upstream chart drift, and how to handle unrelated drift
     (split into a catch-up PR vs land it together)

Surfaced as an unrelated diff in NVIDIA#872 (aws-efa v0.5.26 bump) when
`make bom-docs` was rerun against the rebased branch. Splitting this
catch-up out so NVIDIA#872's diff is scoped to aws-efa content.

Open question for follow-up (not blocking this PR): how did NVIDIA#866 merge
with a stale BOM? `make qualify` enforces BOM-up-to-dateness locally,
so either the CI check is path-conditional and skipped on that PR, or
there's another loophole. Worth a separate audit.

Test plan:
- `make qualify` passes (Go tests, golangci-lint + yamllint, BOM
  staleness check, agents-sync check, chart-pin verification, 20/20
  chainsaw, vulnerability scan, license headers)
- No code changes, no chart pins moved, no behavior change

Two coupled changes:

1. Pure BOM regen of docs/user/container-images.md via `make bom-docs`
   to reflect current registry + chart state. No code changes, no
   registry pins moved; this picks up drift accumulated since main's
   last BOM commit:
   - slinky-slurm-operator and slinky-slurm-operator-crds BOM rows
     added (registry entries landed in NVIDIA#866 without a corresponding
     BOM regen)
   - kubeflow-trainer's bundled pytorch image bumped upstream
   - busybox bumped upstream in some chart's rendered templates

2. Add the BOM-regen rule to contributor and agent docs, surfaced by
   the same NVIDIA#872 review that flagged the drift:
   - .claude/CLAUDE.md (auto-synced to AGENTS.md) — short rule in the
     "Common Tasks" section: after any change to recipes/registry.yaml,
     a component values file, or a chart pin, run `make bom-docs` and
     commit the regenerated container-images.md in the same PR
   - docs/contributor/data.md — new "Regenerating the BOM" subsection
     under Component Configuration covering when to run it, how it can
     surface upstream chart drift, and how to handle unrelated drift
     (split into a catch-up PR vs land it together)

Surfaced as an unrelated diff in NVIDIA#872 (aws-efa v0.5.26 bump) when
`make bom-docs` was rerun against the rebased branch. Splitting this
catch-up out so NVIDIA#872's diff is scoped to aws-efa content.

Open question for follow-up (not blocking this PR): how did NVIDIA#866 merge
with a stale BOM? `make qualify` enforces BOM-up-to-dateness locally,
so either the CI check is path-conditional and skipped on that PR, or
there's another loophole. Worth a separate audit.

Test plan:
- `make qualify` passes (Go tests, golangci-lint + yamllint, BOM
  staleness check, agents-sync check, chart-pin verification, 20/20
  chainsaw, vulnerability scan, license headers)
- No code changes, no chart pins moved, no behavior change

Two coupled changes:

1. Pure BOM regen of docs/user/container-images.md via `make bom-docs`
   to reflect current registry + chart state. No code changes, no
   registry pins moved; this picks up drift accumulated since main's
   last BOM commit:
   - slinky-slurm-operator and slinky-slurm-operator-crds BOM rows
     added (registry entries landed in NVIDIA#866 without a corresponding
     BOM regen)
   - kubeflow-trainer's bundled pytorch image bumped upstream
   - busybox bumped upstream in some chart's rendered templates

2. Add the BOM-regen rule to contributor and agent docs, surfaced by
   the same NVIDIA#872 review that flagged the drift:
   - .claude/CLAUDE.md (auto-synced to AGENTS.md) — short rule in the
     "Common Tasks" section: after any change to recipes/registry.yaml,
     a component values file, or a chart pin, run `make bom-docs` and
     commit the regenerated container-images.md in the same PR
   - docs/contributor/data.md — new "Regenerating the BOM" subsection
     under Component Configuration covering when to run it, how it can
     surface upstream chart drift, and how to handle unrelated drift
     (split into a catch-up PR vs land it together)
   - Makefile — corrected the bom-check target's help text from
     "CI gate, opt-in locally" to "opt-in; not wired into qualify/
     lint/merge gate" to match the actual enforcement story

Surfaced as an unrelated diff in NVIDIA#872 (aws-efa v0.5.26 bump) when
`make bom-docs` was rerun against the rebased branch. Splitting this
catch-up out so NVIDIA#872's diff is scoped to aws-efa content.

Honest enforcement story: an earlier revision of these docs (and an
earlier revision of this commit message) claimed `make qualify` and
CI catch stale BOMs. That was incorrect — verified directly against
the Makefile:
- `qualify` depends on test-coverage / lint / e2e / scan / license-
  check, none of which run `bom-check`
- `lint` runs `bom-pinning-check` (chart-pin verification per ADR-006),
  not `bom-check` (BOM doc freshness)
- The merge gate has no PR-time BOM-staleness check
The corrected docs say so explicitly. Wiring `bom-check` into the gate
is a desirable follow-up but intentionally out of scope here — it
would change CI behavior for every PR and likely block unrelated PRs
that happen to expose accumulated upstream drift, which deserves its
own discussion.

This also explains how NVIDIA#866 merged with a stale BOM: nothing in the
existing gate would have caught it.

Test plan:
- `make qualify` passes (Go tests, golangci-lint + yamllint, agents-
  sync check, chart-pin verification, 20/20 chainsaw, vulnerability
  scan, license headers)
- No code changes, no chart pins moved, no behavior change

…g note

Two follow-ups to PR NVIDIA#866 (Slinky slurm-operator), bundled into one
post-merge cleanup PR.

Platform-enum drift backfill — surfaces NVIDIA#866 did not catch
============================================================

PR NVIDIA#866 added `--platform slurm` to the OpenAPI spec, the Go enum
(GetCriteriaPlatformTypes), and docs/user/{cli,api}-reference.md +
component-catalog.md. Per the project's enum-audit rule (CLAUDE.md
"Documentation updates"), every surface that enumerates platform
values should be updated. The following surfaces still showed
`kubeflow` only (pre-existing drift predating slurm — `dynamo` and
`nim` were already missing):

- docs/README.md:11 — glossary Criteria row
- docs/contributor/data.md:130 — platform row in criteria table
- docs/contributor/validations.md:99 — platform bullet
- site/docs/getting-started/index.md:19 — glossary mirror
- pkg/recipe/criteria.go:246 — Platform field godoc
- pkg/recipe/doc.go:32 — Criteria struct shape comment (was
  non-alphabetical "kubeflow, dynamo, nim, slurm, any")
- pkg/recipe/doc.go:93-98 — CriteriaPlatform* constant listing
  (was non-alphabetical order)
- pkg/api/doc.go:72 — REST API doc was missing slurm entirely
  (only one of these with factual content drift; rest are ordering)
- .claude/skills/analyzing-snapshots/SKILL.md:278 — internal AICR
  snapshot-analysis skill criteria table

All now list `dynamo, kubeflow, nim, slurm` in alphabetical order
matching pkg/recipe.GetCriteriaPlatformTypes() (the authoritative Go
enum). Per-file delimiter convention preserved: slash-separated in
glossary-style parens, comma-separated in tables/godoc.

New guard test
==============

pkg/recipe/doc_test.go::TestCriteriaPlatformConstantsMatchGetter
asserts the CriteriaPlatform* constants and GetCriteriaPlatformTypes()
stay in sync. Catches the exact class of drift this commit fixes if
a future platform value is added to one but not the other.

Catalog node-selector limitation note
=====================================

Appends a `**Known limitation:**` clause to the slinky-slurm-operator
row in docs/user/component-catalog.md, documenting that chart v1.1.0
silently ignores `operator.nodeSelector` and `webhook.nodeSelector`
and linking SlinkyProject/slurm-operator#187 for the upstream fix.
The same limitation is already in-line in recipes/registry.yaml; the
catalog note surfaces it to readers of the rendered user docs.

Originally planned but no longer needed
=======================================

While auditing, I had also planned to add Mark's two "optional nits"
from NVIDIA#866 — a chart-default verification pointer at defaultVersion in
recipes/registry.yaml, and a vacuous-pass guard comment in
recipes/checks/slinky-slurm-operator/health-check.yaml. Both turned
out to already be in the merged version (Mark applied them himself
before merge — see recipes/registry.yaml:550-555 and
recipes/checks/slinky-slurm-operator/health-check.yaml:64-69).

The .lychee.toml header change is the project's license-header check
auto-applying the standard Apache 2.0 copyright block.

Addresses
=========

Mark (@mchmarny) review feedback on NVIDIA#866:
  - Doc-audit gap across user docs (cli-reference + api-reference
    landed at merge time; this PR catches the remaining surfaces)

CodeRabbit review feedback on NVIDIA#866:
  - NodeSelector limitation note on slinky-slurm-operator catalog row

Internal panel review (PE + QA + DA) on draft NVIDIA#884:
  - Extended audit to pkg/api/doc.go (factual miss), pkg/recipe/doc.go
    (ordering), .claude/skills/analyzing-snapshots/SKILL.md
  - Added guard test against future enum/getter drift
  - Squashed to single commit; rebased onto upstream/main HEAD

Signed-off-by: Carlos Eduardo Arango Gutierrez <[email protected]>

…g note

Follow-up to PR NVIDIA#866 (Slinky slurm-operator). Backfills platform-enum
drift on surfaces the original PR did not catch, plus surfaces the
chart v1.1.0 nodeSelector silent-ignore limitation on the user-facing
component-catalog row.

PR NVIDIA#896 (yuanchen8911) already fixed docs/README.md and
docs/contributor/validations.md as part of a broader docs audit. PR
NVIDIA#893 removed the site/docs/ vitepress mirror entirely. Remaining
surfaces this PR addresses:

- docs/contributor/data.md:130 — platform row in criteria table
- docs/user/component-catalog.md:36 — slinky-slurm-operator row,
  appended **Known limitation:** clause documenting chart v1.1.0
  silent-ignore of operator.nodeSelector/webhook.nodeSelector
  (the same limitation is already inline in recipes/registry.yaml;
  this surfaces it on the rendered user catalog)
- pkg/api/doc.go:72 — REST API godoc was missing slurm entirely
- pkg/recipe/doc.go:32 — struct shape comment reordered to alphabetical
  to match GetCriteriaPlatformTypes()
- pkg/recipe/doc.go:93-98 — CriteriaPlatform* constants reordered
  alphabetical
- pkg/recipe/criteria.go:246 — Platform-field godoc updated
- .claude/skills/analyzing-snapshots/SKILL.md:278 — internal AICR
  snapshot-analysis skill criteria table

All now list 'dynamo, kubeflow, nim, slurm' alphabetically matching
pkg/recipe.GetCriteriaPlatformTypes() (the authoritative Go enum).

New guard test
==============

pkg/recipe/doc_test.go::TestCriteriaPlatformConstantsMatchGetter
asserts the CriteriaPlatform* constants and GetCriteriaPlatformTypes()
stay in sync. Mechanically catches the exact class of drift this
commit fixes if a future platform value is added to one but not the
other.

Addressed reviews
=================

@mchmarny (NVIDIA#866) — Doc-audit gap. cli-reference and api-reference were
fixed at merge time and NVIDIA#896 picked up README/validations; this PR
catches the remaining surfaces (contributor/data, the three Go godoc
files, and the internal skill table).

@coderabbitai (NVIDIA#866) — NodeSelector limitation note on the
slinky-slurm-operator catalog row.

Internal PE + QA + DA panel review on draft NVIDIA#884 — extended audit
to pkg/api/doc.go (factual miss), pkg/recipe/doc.go (ordering),
.claude/skills/analyzing-snapshots/SKILL.md, plus the guard test
in pkg/recipe/doc_test.go.

Signed-off-by: Carlos Eduardo Arango Gutierrez <[email protected]>

Add `runai` as a platform value in AICR recipes, peer to
`dynamo`, `kubeflow`, `nim`, and `slurm`.

Mirrors the precedent set by NVIDIA#866 (slurm). Sites updated:

- pkg/recipe/criteria.go: CriteriaPlatformRunai const, parser case,
  sorted slice in GetCriteriaPlatformTypes
- pkg/recipe/criteria_test.go: table-driven cases for runai (lower +
  uppercase) and updated TestGetCriteriaPlatformTypes expectations
- pkg/recipe/doc.go: Platform field docstring + bullet under "Platform
  types for workload frameworks"
- api/aicr/v1/server.yaml: three platform enum sites
- docs/user/{api,cli}-reference.md: platform value enumerations
- docs/contributor/{api-server,validations}.md: platform value
  enumerations
- docs/README.md: glossary Criteria row

Run:ai recipe matrix and any operator/CRD plumbing are deferred to
follow-up PRs.

Per review feedback on NVIDIA#955: pkg/api/doc.go:72 still listed the
platform enum as (dynamo, kubeflow, nim, any) — missing both `slurm`
(pre-existing drift from NVIDIA#866) and the new `runai` added in this PR.

Updated to match the canonical list now used across server.yaml,
pkg/recipe/criteria.go, and the user/contributor docs:

    platform: Platform/framework (dynamo, kubeflow, nim, runai, slurm, any)

No behavioral change; godoc-only.

Reported-by: @yuanchen8911
Signed-off-by: Rob Esker <[email protected]>

Follow-up to PR #866 (Slinky slurm-operator). Backfills platform-enum
drift on surfaces the original PR did not catch, plus surfaces the
chart v1.1.0 nodeSelector silent-ignore limitation on the user-facing
component-catalog row.

PR #896 (yuanchen8911) already fixed docs/README.md and
docs/contributor/validations.md as part of a broader docs audit. PR
surfaces this PR addresses:

- docs/contributor/data.md:130 — platform row in criteria table
- docs/user/component-catalog.md:36 — slinky-slurm-operator row,
  appended **Known limitation:** clause documenting chart v1.1.0
  silent-ignore of operator.nodeSelector/webhook.nodeSelector
  (the same limitation is already inline in recipes/registry.yaml;
  this surfaces it on the rendered user catalog)
- pkg/api/doc.go:72 — REST API godoc was missing slurm entirely
- pkg/recipe/doc.go:32 — struct shape comment reordered to alphabetical
  to match GetCriteriaPlatformTypes()
- pkg/recipe/doc.go:93-98 — CriteriaPlatform* constants reordered
  alphabetical
- pkg/recipe/criteria.go:246 — Platform-field godoc updated
- .claude/skills/analyzing-snapshots/SKILL.md:278 — internal AICR
  snapshot-analysis skill criteria table

All now list 'dynamo, kubeflow, nim, slurm' alphabetically matching
pkg/recipe.GetCriteriaPlatformTypes() (the authoritative Go enum).

New guard test
==============

pkg/recipe/doc_test.go::TestCriteriaPlatformConstantsMatchGetter
asserts the CriteriaPlatform* constants and GetCriteriaPlatformTypes()
stay in sync. Mechanically catches the exact class of drift this
commit fixes if a future platform value is added to one but not the
other.

Addressed reviews
=================

@mchmarny (#866) — Doc-audit gap. cli-reference and api-reference were
fixed at merge time and #896 picked up README/validations; this PR
catches the remaining surfaces (contributor/data, the three Go godoc
files, and the internal skill table).

@coderabbitai (#866) — NodeSelector limitation note on the
slinky-slurm-operator catalog row.

Internal PE + QA + DA panel review on draft #884 — extended audit
to pkg/api/doc.go (factual miss), pkg/recipe/doc.go (ordering),
.claude/skills/analyzing-snapshots/SKILL.md, plus the guard test
in pkg/recipe/doc_test.go.

Signed-off-by: Carlos Eduardo Arango Gutierrez <[email protected]>

…#884)

Signed-off-by: Carlos Eduardo Arango Gutierrez <[email protected]>
Co-authored-by: Mark Chmarny <[email protected]>