3 files changed, 16 insertions, 15 deletions
diff --git a/pi/agent/extensions/agent-plan-mode/README.md b/pi/agent/extensions/agent-plan-mode/README.md
index 4ba700a..5ea23a1 100644
--- a/pi/agent/extensions/agent-plan-mode/README.md
+++ b/pi/agent/extensions/agent-plan-mode/README.md
@@ -26,7 +26,7 @@ todo list.
 - `/task-update <selector> :: <new description>`
   Replace a task description.
 - `/task-modify <selector> :: <mods>`
-  Apply `~/go/bin/do modify` arguments to a task.
+  Apply `~/go/bin/ask modify` arguments to a task.
 - `/tasks`
   Show started and `+READY` tasks for the current repo.
 - `/task-next [run]`
@@ -40,10 +40,10 @@ todo list.
 
 ## Rules
 
-- all task operations go through `~/go/bin/do`, never raw `task`
-- tasks are scoped to the current git repo through the `~/go/bin/do` wrapper
+- all task operations go through `~/go/bin/ask`, never raw `task`
+- tasks are scoped to the current git repo through the `~/go/bin/ask` wrapper
 - use alias IDs for task references
-- `~/go/bin/do add` prints `created task <alias-id>`, and subsequent task operations should keep using that alias ID
+- `~/go/bin/ask add` prints `created task <alias-id>`, and subsequent task operations should keep using that alias ID
 - planning mode is read-only by design
 - the extracted plan is session-local, so `/plan`, the planning prompt,
   `/plan-create-tasks`, and `/plan-exit` should happen in the same interactive
@@ -160,13 +160,13 @@ Analyze the repo and give me a Plan: for the next implementation slice.
 ## Notes And Limits
 
 - Planning mode is read-only by design.
-- All task operations still go through `~/go/bin/do`, never raw `task`.
-- `~/go/bin/do` uses subcommand syntax. It is not a natural-language
-  task assistant and should never be called like `~/go/bin/do agent-task-management ...`.
+- All task operations still go through `~/go/bin/ask`, never raw `task`.
+- `~/go/bin/ask` uses subcommand syntax. It is not a natural-language
+  task assistant and should never be called like `~/go/bin/ask agent-task-management ...`.
 - Execution mode injects the current task back into the agent prompt
   so the model works against the real task rather than an in-memory checklist.
 - Execution mode treats the focused task as the already-selected starting
-  point and blocks repeated identical `~/go/bin/do info <id>` lookups until the
+  point and blocks repeated identical `~/go/bin/ask info <id>` lookups until the
   agent has moved on to repo inspection, implementation, tests, review, or a
   different command.
 - Full `/plan` state is not meant to be passed across unrelated one-shot `pi -p`
diff --git a/pi/agent/extensions/agent-plan-mode/utils.ts b/pi/agent/extensions/agent-plan-mode/utils.ts
index 3aeccbf..f388dff 100644
--- a/pi/agent/extensions/agent-plan-mode/utils.ts
+++ b/pi/agent/extensions/agent-plan-mode/utils.ts
@@ -1,13 +1,14 @@
 /** Shown in prompts/docs for Pi skills and extensions (task CLI). */
-export const DO_CLI_REF = "~/go/bin/do";
+export const DO_CLI_REF = "~/go/bin/ask";
 
 export function resolveDoExecutable(): string {
-	return `${process.env.HOME ?? ""}/go/bin/do`;
+	return `${process.env.HOME ?? ""}/go/bin/ask`;
 }
 
-const DO_INVOCATION_PREFIXES = ["~/go/bin/do", "do"] as const;
+/** Full path and bare name for `ask`; legacy `do` prefixes still match old prompts/scripts. */
+const DO_INVOCATION_PREFIXES = ["~/go/bin/ask", "ask", "~/go/bin/do", "do"] as const;
 
-/** Bash may use `do` or the full `~/go/bin/do` path. */
+/** Recognize task CLI invocations (prefer `ask`; `do` kept for backward compatibility). */
 export function matchDoInvocation(trimmed: string): { rest: string } | undefined {
 	for (const prefix of DO_INVOCATION_PREFIXES) {
 		if (trimmed === prefix) return { rest: "" };
diff --git a/pi/agent/extensions/nemotron-tool-repair/index.ts b/pi/agent/extensions/nemotron-tool-repair/index.ts
index 0fb868c..3eb1e87 100644
--- a/pi/agent/extensions/nemotron-tool-repair/index.ts
+++ b/pi/agent/extensions/nemotron-tool-repair/index.ts
@@ -36,15 +36,15 @@ Additional tool-use discipline for this model:
 - Do not emit example tool-call markup or pseudo-tool syntax for the user to read.
 - Emit at most one tool invocation at a time, then wait for the tool result.
 - After a tool result, continue from that result instead of restating the plan.
-- There is no "submit" tool. When your task is complete, respond with your final answer directly. If you are working on a tracked task, mark it done via bash: ~/go/bin/do done <id>.
+- There is no "submit" tool. When your task is complete, respond with your final answer directly. If you are working on a tracked task, mark it done via bash: ~/go/bin/ask done <id>.
 `.trim();
 
 // Qwen Coder models are trained on agent frameworks that include a "submit" tool
 // as a task-completion signal. Since no such tool exists here, calling it causes
-// a "Tool submit not found" error. Redirect to a final answer or ~/go/bin/do done instead.
+// a "Tool submit not found" error. Redirect to a final answer or ~/go/bin/ask done instead.
 const QWEN_TOOL_DISCIPLINE = `
 Additional tool-use discipline for this model:
-- There is no "submit" tool. When your task is complete, respond with your final answer directly. If you are working on a tracked task, mark it done via bash: ~/go/bin/do done <id>.
+- There is no "submit" tool. When your task is complete, respond with your final answer directly. If you are working on a tracked task, mark it done via bash: ~/go/bin/ask done <id>.
 `.trim();
 
 interface FileModelConfig {