sync: auto-sync from GURU-5070 at 2026-07-01 13:06:10

Author: Mike Swanson
Machine: GURU-5070
Timestamp: 2026-07-01 13:06:10
This commit is contained in:
2026-07-01 13:07:04 -07:00
parent 29355584bf
commit af8a3de00e
5 changed files with 330 additions and 1 deletions

View File

@@ -34,6 +34,7 @@
- [GuruRMM technical reference](reference_gururmm.md) — Server (172.16.3.30) layout + downloads dir `/var/www/gururmm/downloads` + `.channel` sidecar rollout control (stable/beta) + privileged server access via the server's OWN root RMM agent (hostname `gururmm`, no SSH needed; plink fallback) + API + `context=user_session` (WTS impersonation) + build-pipeline vendoring at `deploy/build-pipeline/` + Linux agent systemd sandbox trap.
- [GuruRMM command timeout_seconds](reference_gururmm_command_timeout_seconds.md) — agent command dispatch honors `timeout_seconds`, NOT `timeout`; long jobs die ~300s / go zombie (`running`, empty stdout) otherwise. Cost Birth Biologic a full day.
- [SharePoint Graph large-file upload](reference_sharepoint_graph_large_file_upload.md) — <4MB simple PUT, >=4MB MUST use chunked upload session (Content-Range); `\\?\` long paths; idempotent size-check; verify counts via /root/delta; single stream ~40Mbps (SPO throttle).
- [RMM-spawn headless Claude](reference_rmm_spawn_headless_claude.md) — run `claude -p` on any RMM-managed Windows box with Claude Code (reaches coord-isolated sites like AD2); use `context:user_session`, UNSET the stale machine `ANTHROPIC_API_KEY` (shadows OAuth → "Invalid API key"), detach + poll a DONE marker. Validated on AD2 2026-07-01.
- [RMM agent update model](rmm-agent-update-model.md) — Agent updates are server-PUSH on heartbeat (no self-poll); available versions = filesystem scan needing a `.sha256`; promote flips `.channel` sidecars beta→stable globally. Two stranders: beta-first freezes stable until an explicit promote; agents older than ~0.6.50 re-enroll with a NEW device_id/agent row when updated.
- [GuruRMM physical server storage](gururmm-physical-server-storage.md) — New box 172.16.1.231 (temp IP→will be .30), Ubuntu 26.04, ssh key `gururmm-physical`/alias `gururmm-new`. SSD (915G root) = HOT (PG default tablespace + WAL + builds); HDD ext4 at `/data` = COLD (`gururmm_cold` PG tablespace for aged `agent_logs` partitions + downloads + backups + archive). The #3 retention answer.
- [Trebesch DESKTOP-QNP3ON5 shell replacement](reference_trebesch_qnp3on5.md) — AT Trebesch box runs an Explorer shell replacement; explorer.exe owner check returns blank — use Win32_ComputerSystem.UserName. GuruRMM SWIFT-LION-2892.
@@ -168,6 +169,7 @@
- [Dataforth history (2026-03-27 incident)](project_dataforth_history.md) — DF-JOEL2 compromise via ScreenConnect social-engineering, attacker C2 IPs + IC3 case + remediation log + MFA rollout origin story + Joel Lohr retirement. RESOLVED 2026-04-04.
- [Radio show co-host — Tara, not Tom](radio_show_no_cohost_named_tom.md) — Co-host in 2014-s6e19 and 2016-s8e43 is Tara. "Tom" was hallucinated; rename complete.
- [Proposal: centralize config in identity.json](proposal_identity_centralization.md) — Rationale for the identity.json machine-config centralization (claudetools_root, ollama/python); now implemented.
- [Defender: exempt all Claude activity](feedback_defender_claude_exclusions.md) — Mike wants NOTHING Claude/ClaudeTools issues flagged by Defender; ClickFix threat IDs 2147939088/2147945138 (RMM curl dispatch AMSI FPs) set to Allow + broad process/path exclusions. ThreatID-Allow is the lever; exclusions alone don't stop AMSI CmdLine detections.
- [ACG MSP tool stack](reference_acg_msp_stack.md) — ScreenConnect/CW Control, Splashtop, Syncro, Datto RMM, Datto EDR/AV, GuruRMM are ACG's OWN tools; do not flag as foreign/threat on managed machines (Defender-off is expected when Datto AV is active).
- [VoIP vendor stack: PacketDial / OIT / NetSapiens / YMCS](reference_packetdial_oit_netsapiens.md) — PacketDial = ACG's VoIP-dept brand (pbx.packetdial.com, the `packetdial` skill); NetSapiens = the PBX platform (API v2); OIT/OITVOIP = white-label wholesaler running NetSapiens (api.ucaasnetwork.com); reseller `91912.service`. YMCS (Yealink) = phone device-mgmt, pairs with the PBX.
- [ACG Website Hosting](project_azcomputerguru_hosting.md) — azcomputerguru.com is hosted on IX Web Hosting via cPanel.
@@ -200,4 +202,4 @@
- [GuruScan verification IN TEST / paused](project_guruscan_in_test_paused.md) — multi-engine scanner verify on DESKTOP-MS42HNC paused 2026-06-22 (VM rebooted mid-Emsisoft run); HitmanPro done (36 removed), Emsisoft full-scan unverified; resume `guruscan-agent-test.sh DESKTOP-MS42HNC scan-one Emsisoft`; Defender RTP/Tamper still off on VM
- [GuruRMM fleet dispatch-hang fix](project_gururmm_dispatch_hang_fix.md) — blocking send_to on a full bounded channel to one black-holed agent wedged ALL command dispatch; fixed with try_send (9dae20c, deployed); proper black-hole eviction still missing (was reverted in 80df458) — finish it if it recurs
- [Windows won't-boot / offline DISM repair playbook](windows-offline-dism-repair-gotchas.md) — Automatic Repair loop = boot-critical fault (disk/registry/wedged update), NOT shell/appx store corruption (that's a symptom); `FaultyPackageInProgress` + 100s of Install/Uninstall-Pending packages = wedged CU -> RevertPendingActions or clean install. Offline DISM rejects `wim:` source (0x800f082e) -> MOUNT the wim, source `\Windows`. Ventoy breaks WIM mount (0xc1420134) -> use Rufus. 25H2(26200)=24H2(26100)+enablement, so match 26100 media. First hit: Four Paws AvImark #32447.
- [Remediation-tool has full M365 access (incl. SharePoint)](reference_remediation_tool_365_access.md) — the app suite covers Graph/EXO/Defender/SharePoint; don't declare "no access" on an accessDenied. SharePoint app-only needs a CERT (secret = "Unsupported app only token"); use get-token.sh `sharepoint`/`sharepoint-admin` tiers + CSOM admin API (Graph /admin/sharepoint/settings scope not held). Full map: skill references/app-permissions-and-sharepoint.md.
- [Remediation-tool has full M365 access (incl. SharePoint)](reference_remediation_tool_365_access.md) — the app suite covers Graph/EXO/Defender/SharePoint; don't declare "no access" on an accessDenied. SharePoint app-only needs a CERT (secret = "Unsupported app only token"); use get-token.sh `sharepoint`/`sharepoint-admin` tiers + CSOM admin API (Graph /admin/sharepoint/settings scope not held). Full map: skill references/app-permissions-and-sharepoint.md.

View File

@@ -0,0 +1,18 @@
---
name: feedback_defender_claude_exclusions
description: Mike wants NOTHING Claude/ClaudeTools issues to be flagged by Windows Defender; keep broad exclusions + allow the ClickFix threat IDs that fire on RMM curl dispatch.
metadata:
type: feedback
---
On his workstation (GURU-*), Mike wants **nothing Claude issues to be affected by Defender AV** — it's a constant irritation. The recurring hits are `Trojan:Win32/ClickFix.DBD!MTB` (ThreatID 2147939088) and `Trojan:Win32/ClickFix.ZF` (ThreatID 2147945138), fired by Defender's AMSI **command-line** scan on the base64-PowerShell payloads that `curl.exe` POSTs to the GuruRMM coordination API (`172.16.3.30:3001/api/agents/.../command`).
**Why:** These are false positives on legitimate ClaudeTools/GuruRMM command dispatch. He's the admin/owner and made an informed call to allow the family.
**How to apply:** Process/path exclusions alone do NOT stop these — AMSI CmdLine/behavioral detections ignore `ExclusionProcess`/`ExclusionPath`. The lever that works is `Add-MpPreference -ThreatIDDefaultAction_Ids <id> -ThreatIDDefaultAction_Actions Allow` (Allow = action 6) for both IDs. Also maintained (elevated PowerShell):
- ExclusionProcess: bash.exe, curl.exe, git.exe, node.exe, claude.exe
- ExclusionPath: `C:\Program Files\Git` (+ mingw64\bin, usr\bin), `C:\Program Files\nodejs`, `C:\Users\<u>\.claude`, `C:\Users\<u>\.local\bin`, `C:\Users\<u>\AppData\Roaming\npm`, `C:\ClaudeTools`, `D:\ClaudeTools`.
**ACTIVE (2026-07-01):** Mike opted for the fully-blanket lever — `Set-MpPreference -DisableScriptScanning $true` is SET on this box, disabling Defender AMSI script scanning machine-wide (his call: "I'm not likely to fall for bogus scripts"). This alone stops the CmdLine detections regardless of variant ID; the ThreatID-Allows + exclusions remain as belt-and-suspenders. If ever re-enabling, `Set-MpPreference -DisableScriptScanning $false`.
**Fleet application (2026-07-01):** `DisableScriptScanning` is **Tamper-Protection-gated** — it silently stays `False` if TP is on, even from SYSTEM. This workstation's TP is OFF (toggle worked); **GURU-BEAST-ROG's TP is ON**, so on Beast only the exclusions + ClickFix ThreatID-Allows applied via RMM (those aren't tamper-gated and DO cover the recurring detections) — the blanket script-scanning kill there needs a manual Windows Security UI toggle (TP can't be disabled by script). Beast (GURU-BEAST-ROG, AZ Computer Guru/Mike's House, RMM id 5233d75b-...) is "treated like this machine." Howard was OFFERED the same via Discord DM — his choice on his own box; do NOT push to Howard's machine without his ok. Related: [[reference_acg_msp_stack]] (ACG's own tools shouldn't be flagged as threats), [[feedback_windows_quote_stripping]].

View File

@@ -0,0 +1,33 @@
---
name: rmm-spawn-headless-claude
description: Spawn a headless `claude -p` on any RMM-managed Windows box that has Claude Code installed — reaches isolated sites (AD2) the coord API can't
metadata:
type: reference
---
Any RMM-managed Windows endpoint with Claude Code installed can run an autonomous headless
Claude, dispatched via a GuruRMM command — even a site that's isolated from the ACG coord API.
The RMM agent phones home outbound, so this works where [[ad2-comms-via-sync-only]] says coord
can't reach (coord `:8001` blocked ≠ RMM `:3001` blocked). Validated 2026-07-01 on AD2
(Dataforth DC, agent `cfa93bb6-...`, claude v2.1.181 at `C:\Users\sysadmin\.local\bin\claude.exe`).
Recipe:
- Dispatch with **`"context":"user_session"`** — needs an interactive logged-on user (check
`quser`); an admin session comes back elevated. `claude` is a per-user install, not on the
SYSTEM PATH, so SYSTEM context won't find it.
- **GOTCHA: unset `ANTHROPIC_API_KEY` first.** A stale machine-level `ANTHROPIC_API_KEY` (108-char)
shadows the good OAuth creds and makes `claude -p` fail with `Invalid API key · Fix external API
key`. `Remove-Item Env:\ANTHROPIC_API_KEY` (+ `$env:ANTHROPIC_API_KEY=$null`) before invoking →
falls back to `~\.claude\.credentials.json` OAuth and authenticates.
- **Detach + poll.** A real audit run takes many minutes; RMM caps command lifetime (see
[[gururmm-command-timeout-seconds]] — use `timeout_seconds`). Launch detached
(`Start-Process powershell -File runner.ps1 -WindowStyle Hidden`), have the runner write the
deliverable to a file + a `DONE.txt` marker, and poll the marker via short RMM commands.
- Run headless as: `claude -p <brief> --permission-mode bypassPermissions --output-format text`.
For an audit, give an ironclad READ-ONLY brief (no writes/git/state changes) since
bypassPermissions lets it run any tool. Pass the brief via a base64'd file to dodge quoting.
- Windows/Git-Bash: the mingw `curl` intermittently hits `Permission denied` (AV lock) —
use `/c/Windows/System32/curl.exe` for the dispatch. See [[feedback_windows_quote_stripping]].
Use for: live audits/data-gathering on isolated or hard-to-reach managed boxes without the async
sync-handoff. Keep it read-only on production (AD2 is a domain controller).