syncro: expand skill to full Syncro API surface (assets/RMM/deploy) + verified capability model

Add comprehensive Syncro coverage beyond PSA core:
- New .claude/standards/syncro/api-reference.md: complete verified inventory of ~180
  endpoints across 38 resource types (generated from live OpenAPI 3.0 spec + tenant
  probe 2026-07-10), with worked GET/POST/PUT/DELETE templates and token-capability matrix.
- /syncro: asset read intelligence (patches, installed_applications), asset create/update,
  policy-folder move (move-asset), RMM alerts, and deploy-agent (hybrid installer push via
  GuruRMM using SyncroSetup --console --allow-force-reboot).
- move-asset ships a capability preflight (GET /policy_folders?customer_id -> 401 = missing
  Assets-Policy-Change) + mandatory post-write verify, because an under-scoped token returns
  HTTP 200 and silently no-ops the move.

Correct the "Syncro RMM is API-impossible" belief: it was a token-scope gap, not an API
limit. Live-verified the asset move (flip-and-restore 692253->692278->692253). Token scope
today: Howard + Winter full; Mike (vaulted ...ebbeb3) still 401 pending re-vault.

Corrects memory reference-syncro-rmm-api-gui-only; correction logged to errorlog.md.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-07-10 15:45:26 -07:00
parent 46d2ae7a5b
commit 1f56098652
5 changed files with 629 additions and 26 deletions

View File

@@ -19,6 +19,8 @@ Categories (the `[type]` tag): _(none)_ = skill/command execution failure ·
<!-- Append entries below this line -->
2026-07-10 | Howard-Home | syncro/assets | [correction] assumed asset writes (POST/PUT create-update), policy_folder_id move, patches, and installed_applications were not doable via the public API; correct is the endpoints exist per the live OpenAPI spec and the prior 'policy_folder_id silently ignored' was a token missing the 'Assets - Policy Change' scope, not an API limitation [ctx: ref=reference-syncro-rmm-api-gui-only endpoint=PUT_/customer_assets/{id}]
2026-07-10 | GURU-BEAST-ROG | guruscan | detached full-chain wrapper died mid-RKill on TPM-PC (marker never written, task->Ready, rkill.log completed fine as orphan) — RKill's process sweep almost certainly killed its own hidden SYSTEM powershell parent; relaunched with -SkipScanners RKill [ctx: host=TPM-PC client=mineralogical-record chain-design-issue]
2026-07-10 | GURU-5070 | packetdial | HTTP 404 GET https://pbx.packetdial.com/ns-api/v2/domains/vwp.91912.service/users/100: {"code":404,"message":"Resource not found."} [ctx: cmd=user] (x2)