Slash commands

All commands run inside a Mythic chat channel using the Sage model (or Sage Watcher, where noted).

Command	What it does
`/state`	Show the engagement ledger, the durable hop table of proven effects shared by chats in the current Mythic operation engagement
`/state objective <text>`	Set the engagement objective
`/state reconcile [task_id] [apply]`	Import verified effects and credentials from completed Mythic tasks (dry-run unless `apply` is given, because task output is attacker-influenceable)
`/state set <row> <status>` · `/state remove <row>` · `/state wipe`	Edit or clear ledger rows (you cannot hand-promote a row to `achieved`; that needs a real proof envelope)
`/mode`	Show the channel’s current mode
`/mode conversation\|supervised\|auto`	Switch the channel’s mode
`/list`	List active Sage chat sessions
`/tools`	List Sage’s public Mythic tool functions, descriptions, and access requirements: a deterministic inventory, no model call
`/report [identity]`	Show retained identity/membership observations shared across assisted chats in this operation engagement, unknown access prerequisites, and next checks: one bounded check, no new task
`/stop`	Cooperatively stop the running agent on this channel
`/bloodhound [force] [dir]`	Connect, report, or rebind the BloodHound MCP; see BloodHound
`/mcp <list\|tools\|call\|connect\|disconnect\|policy>`	Manage MCP servers; see Connecting MCP servers
`/sandbox [shell\|python] <code>`	Run an isolated local snippet (requires the `callback.write` scope)

Watcher commands

Available with the separate Sage Watcher model; see the findings watcher:

Command	What it does
`/watcher apply`	Claim the operation’s sole Watcher owner generation (from the locked `Sage Watcher` channel)
`/watcher status`	Inspect the redacted owner, generation, route sources, cadence, and health
`/watcher scan`, `/watcher pause`, `/watcher resume`, `/watcher interval`	Owner-only scheduler controls

#sage-findings is a normal Mythic channel, not a command: the Watcher posts full finding updates there, and /findings (aliased by the legacy /finding spelling) shows the same canonical view plus watcher health from either model. The bare word findings in conversation is an ordinary prompt with no special behavior.
/state is available in conversation, supervised, and auto modes. Its ledger is shared across chats in the same operation engagement and persists once written. A new engagement can show an empty ledger: viewing operation details or callbacks does not record an achieved effect, and reading /state alone does not create a ledger file.
/tools and simple questions such as “list your tools” use the same deterministic inventory without a model call. Before a session is initialized, live token availability is unverified.