StackJack 0.1.0.10433+e572b66
StackJack 0.1.0.10433+e572b66
Azure run-command and effective network reads now return their actual results instead of an empty response. The CIPP business email compromise check can return only the report sections you ask for. Several fixes cover Halo roles, catalog mode switching, and reconnecting a connector right after removing it, and two portal pages load faster.
New features
- Azure VM run command returns its output.
azure_run_vm_commandnow waits for the script to finish and returns its standard output and error. An optional wait time (0–180 seconds, default 45) controls how long it waits. If the script is still running when the wait ends, the tool returns Azure's current operation status and a way to read the result later withazure_raw_get. The tool's guidance now warns agents not to call it again to fetch output, because that would run the script a second time. - Azure effective routes and effective NSG rules return their data.
azure_list_nic_effective_routesandazure_list_nic_effective_nsg_rulesnow wait for Azure to finish and return the route table or rule set. Previously they returned an empty result. Both accept the same optional wait time. They only read data, so calling them again safely collects a result that took longer than the wait. - Power-state-only VM listing.
azure_list_vmshas a newstatusOnlyoption for a lighter listing focused on VM status. - Filter the CIPP business email compromise report by section.
cipp_bec_checkaccepts an optional list of section names when collecting a finished report. Only those sections are returned, along with the run's identifying details, which makes triage easier when full reports run to many megabytes. If you leave the list out, you get the full report as before. Existing parameters are unchanged.
Improvements
halo_get_rolenow includes the role's permissions by default. You can turn this off with a new argument.- Several Azure tool descriptions are more accurate:
- The resource-group VM list no longer promises power state that Microsoft doesn't return in that form. It now points agents to
azure_list_vmswithstatusOnlyor to the VM instance view tool. - The VM list explains when instance details can be requested.
azure_raw_getdescribes the query strings it accepts.- The
halo_update_roledescription now says which role ID to use.
Optimizations
- The Team page loads per-member usage counts much faster, especially for organizations with a lot of usage history.
- The portal home page loads the client count tile faster, especially for organizations with many configured clients.
Bug fixes
- Halo roles could not be opened or deleted.
halo_get_roleandhalo_delete_rolenow accept the role ID returned byhalo_list_roles, so these tools work as expected. - Reconnecting a connector immediately after removing it could lose the new sign-in. Removing a connection and adding it back within a few seconds could discard the credentials you had just entered. StackJack now waits for the removal to finish and keeps the new sign-in.
- Switching catalog mode could cause the full tool catalog to be served. Some organizations had not chosen a catalog setting, so the compact catalog had been applied automatically. On those connections, asking the AI to change catalog mode could save a preference that caused the full catalog to be served on the next request, and the AI client would then cut it off. The request now saves nothing. If compact mode is requested, the tool confirms it is already in effect. Any other mode returns guidance that an administrator needs to turn on catalog modes in Settings.
- Azure no longer advises a retry after it has accepted a request. Previously, a network problem or server error while StackJack was waiting on an accepted Azure operation was reported as a retryable error, which could lead to a VM script running twice. These cases are now reported as accepted-but-no-longer-followed, with the path to check the result later.
- Azure operation error messages stay accurate. Some error messages could be replaced with unrelated throttling or credential advice. They now keep the correct "do not send again" wording and the path for reading the result later.
- Azure waits now respect the wait time you choose. If Azure asks StackJack to back off for longer than the remaining wait time, StackJack now stops waiting right away. It returns the latest status or the read-later path instead of running past your chosen wait.
- Azure operations that never report a status no longer look finished. If an operation hadn't reported a status before the wait ran out, StackJack returned an empty result, which looked like there was nothing to track. It now returns the read-later path.
- Azure operations that succeeded are no longer reported as in progress. If an operation succeeded but its result couldn't be read before the wait ran out, it is now reported as succeeded rather than still in progress.