sync spec: document the cache values query run actually accepts

[ernestoongaro]

Summary

Headline: omni query run's cache field now documents the values the API really accepts: Standard, SkipRequery, SkipCache, SkipCacheAndRebuildExtracts. The old values (disabled | normal | refresh | refresh_all) came from the spec and always returned a 400, so anyone following omni query run --schema hit that error. timezone is also documented now. Upstream fix: exploreomni/omni#64011.

Also pulled in by the full sync:

• New skills group: list, get, create, update, delete (from exploreomni/omni#62425). Create and update take --body.
• query wait: jobIds is no longer required. The deprecated job_ids is documented and shows up as --job-ids-2, because the generator renames flags that collide.
• models content-validator: new --force-full-validation query param.
• Doc-only schema changes: document/dashboard payloads now have schemas, plus composite topic schemas, AI routines, and the v2 draft app routes.

Test plan

• make build

• make test

• omni query run --schema shows the new cache enum and timezone

• omni skills --help lists the 5 new commands

• omni query wait --help / omni models content-validator --help show the new flags

• Live, read-only on a playground instance:

  • query run with cache set to Standard, SkipRequery, SkipCache, SkipCacheAndRebuildExtracts, and with no cache: all return results
  • query run with the old refresh / normal: 400 from the API, as expected (the API never accepted them)
  • query run with timezone: America/Los_Angeles: works
  • query wait --job-ids <id> and --job-ids-2 <id>: both stream the finished job
  • skills list: returns a paginated response
  • models content-validator-get <model> --force-full-validation true: works

🤖 Generated with Claude Code

https://claude.ai/code/session_01NcBoDwoiTDYfkGVJgy2W3B

[ernestoongaro]

Tested this build against a live instance.

query run cache fix

• cache set to Standard, SkipRequery, SkipCache or SkipCacheAndRebuildExtracts → all return results. The old refresh / normal still return the API's 400 (expected one of "Standard"|"SkipRequery"|"SkipCache"|"SkipCacheAndRebuildExtracts"), and --schema now lists the correct values.
• "timezone": "America/Los_Angeles" → OK

Other changes in this sync

• query wait --job-ids <id> and the deprecated --job-ids-2 <id> → both stream the completed job and footer
• models content-validator-get <model> --force-full-validation true → OK
• models get-topic --schema and -o human still render after the topic response became regular-or-composite
• skills, full lifecycle on a throwaway skill: create → get (JSON and human) → list --identifier → update one field → update --body - from stdin → delete → get returns 404. A duplicate identifier returns 409 and an invalid identifier returns 400; both exit 1 with a single JSON error on stderr.
• AI routines: condition.verify was removed upstream, so the condition is always checked. No CLI changes needed.
• Document and dashboard payloads now have full schemas. Wire format is unchanged.