feat(classic): add computer groups and mobile device groups commands

[neilmartin83]

Summary

Adds Classic API support for /computergroups and /mobiledevicegroups via the classic generator manifest:

• classic-computer-groups (computer_group)
• classic-mobile-device-groups (mobile_device_group)

Both expose full CRUD (list/get/create/update/delete) plus name-based apply and --name lookup.

Why, given modern smart/static group endpoints exist

The classic combined endpoint surfaces the XML group model — smart criteria, static membership, and the *_additions/*_deletions update semantics — which is better suited to manipulating static group membership than the modern split smart/static endpoints. Overlap is intentional. Precedent: advanced computer/mobile searches were added to the classic manifest the same way.

Changes

• specs/classic/resources.yaml — two new resource entries
• internal/commands/groups.go — map both into the Classic Computers / Classic Mobile Devices help groups
• regenerated classic command + registry files (make generate)

Verification

• make lint — 0 issues
• make test — all packages pass
• Live tested against a platform-gateway instance: list (283 computer / 35 mobile groups), get by id, get --name — all return correct XML with criteria

🤖 Generated with Claude Code

[ktn-jamf]

Tip

✅ Merge-ready — Adds Classic API classic-computer-groups and classic-mobile-device-groups commands (full CRUD + name-based apply/--name lookup) via two declarative entries in the classic manifest. No critical or important issues found.

This is a 1029-line diff, but 1006 of those lines are mechanically generated. The actual decision surface is 14 hand-written lines: a 12-line spec addition in specs/classic/resources.yaml and 2 lines of help-group wiring in groups.go. I verified the generated output rather than reviewing it line-by-line.

What I verified:

• No generator drift — make verify-generated passes: the committed generated files (classic_computer_groups.go, classic_mobile_device_groups.go, and the three registry files) exactly match what make generate produces from the new spec. Nothing under generated/ was hand-edited, per the CLAUDE.md boundary.
• Build + tests pass — go build ./... clean; TestApplyProGroups/grouping/classic tests pass. The CLAUDE.md tripwire (TestApplyProGroups_AllCommandsGrouped failing when a new resource isn't grouped) is satisfied because both resources were correctly mapped into the Classic Computers / Classic Mobile Devices help groups.
• Spec correctness — Explicit singular: computer_group / mobile_device_group is necessary and correct; the parser's auto-derivation (TrimSuffix(name,"s")) would otherwise produce the wrong computergroup. Omitting operations/lookups correctly defaults to full CRUD + [id, name], which is what yields apply and --name lookup. The two entries match the cited advancedcomputersearches precedent field-for-field.
• No injection / no conflict — generated paths use url.PathEscape on all user input; get/update --name route through the API's /name/{name} endpoint. The new standalone mobiledevicegroups resource is orthogonal to the pre-existing groups_path: mobiledevicegroups on classic-mobile-devices (that field drives a --group bulk-delete-of-members flag — different command, no name clash).

What's done well

✅ The PR description documents a non-obvious design decision well — it explains why the combined classic /computergroups endpoint is worth adding despite the modern split smart/static group commands (it surfaces the XML group model with *_additions/*_deletions static-membership semantics), cites the advanced-searches precedent, and flags the overlap as intentional. The classic- prefix cleanly avoids collision with the modern smart-computer-groups/static-computer-groups commands. The (smart and static) descriptions are accurate per docs/GLOSSARY.md.

✅ Live-tested against a real platform-gateway instance (283 computer / 35 mobile groups, get by id and --name) — the one thing generated-code review can't fully substitute for.

Review coverage

• Design and architecture: combined classic endpoint, justified overlap with modern split groups, follows advanced-searches precedent, classic- prefix avoids command collision
• Correctness: spec fields correct (explicit singular avoids bad auto-derivation), defaults yield CRUD + name + apply, generated code matches spec (no drift), build + tests pass, name lookup uses API /name/ endpoint
• Security: paths are constant strings; user input passes through url.PathEscape — no injection vector; no auth/secrets surface
• [na] Performance: list is a single GET identical to 50+ other classic resources; no hot paths or queries introduced
• Test coverage: generated commands covered by generator tests + grouping test + smoke registry; author live-tested; no bespoke logic to unit-test
• [na] Reliability: no new error handling/retry logic; inherits internal/client behavior
• Code quality: spec entries match precedent field-for-field; consistent naming and descriptions
• [na] Simplification: minimal by construction (declarative 12-line spec change)
• [na] Frontend concerns: no UI changes
• Project rules compliance: no manual edits to generated/ (verified via no-drift), spec-driven workflow per CLAUDE.md, resources grouped in groups.go

Scope of review

• Reviewed in full: specs/classic/resources.yaml (+12), internal/commands/groups.go (+2), and all 5 generated files (registry/smoke/backup/provenance + spot-check of classic_computer_groups.go).
• Related files checked: generator/classic/parser.go (defaults for operations/lookups/singular), generator/classic/generator.go (groups_path/GroupPath semantics), docs/GLOSSARY.md (smart/static terminology).
• Verification run: make verify-generated, go build ./..., go test ./internal/commands/... -run 'TestApplyProGroups|Grouped|Classic'.
• No prior reviews on this PR.
• Specialist agents: not dispatched — no security/concurrency/performance surface in a declarative spec + generated CRUD wrappers; would have been noise.
• Note on generated Short/Example text ("List all computergroups" uses the raw resource name): template-driven and identical across all 55 classic resources — not flagged, since fixing it means editing a generator template (out of scope) and it's a consistent existing convention.

Generated by pr-review v1.7.0, a Jamf Claude Code skill