prompts: rename task CLI examples from ask to do

Update agent command docs and auditing-code-quality skill so examples
match the Hexai task binary (do, formerly ask).

Made-with: Cursor

3 files changed, 13 insertions, 13 deletions

diff --git a/prompts/commands/create-tasks.md b/prompts/commands/create-tasks.md
index 19161d1..2489577 100644
--- a/prompts/commands/create-tasks.md
+++ b/prompts/commands/create-tasks.md
@@ -60,10 +60,10 @@ You are an assistant that, when invoked via `/create-tasks`, turns a high-level
     - For each task:
       - Use a concise, verb-first description (e.g., "Configure GitHub Actions workflow for tests").
       - Add meaningful tags (e.g., project name, area like `dev`, `docs`, `infra`).
-      - Prefer tying tasks to the **current git project** when invoked inside a repo. Use `ask add` (not `task add`) so the project name is derived automatically from the git root — never type the project name manually.
+      - Prefer tying tasks to the **current git project** when invoked inside a repo. Use `do add` (not `task add`) so the project name is derived automatically from the git root — never type the project name manually.
       - Add an annotation that links back to the plan file path, for example:
         `"See overall plan: <relative/or/absolute/path/to/plan.md>"`.
-    - Use `ask add "<description>" [+tag ...]` to create each task. The `ask` CLI automatically scopes tasks to the current git project, preventing typos in project names.
+    - Use `do add "<description>" [+tag ...]` to create each task. The `do` CLI automatically scopes tasks to the current git project, preventing typos in project names.
 
  4. **Keep steps and tasks small but self-contained**
     - Prefer more, smaller tasks over fewer, very large tasks, as long as each:
diff --git a/prompts/commands/work-on-tasks.md b/prompts/commands/work-on-tasks.md
index 770e5ad..b8d68d1 100644
--- a/prompts/commands/work-on-tasks.md
+++ b/prompts/commands/work-on-tasks.md
@@ -27,31 +27,31 @@ You are the **orchestrator**. You pick tasks, mark them started, launch a sub-ag
 
 1. **Load project-scoped tasks**:
    - Detect the current project from local git context (`git rev-parse --show-toplevel`)
-   - Run `ask ready | head` to list actionable tasks
+   - Run `do ready | head` to list actionable tasks
    - Ignore completed/deleted tasks and non-actionable blocked items
 
 2. **Pick the next task** (default strategy: `{{strategy|highest-impact}}`):
    - Choose one actionable task based on impact, urgency, and clarity
    - If two tasks are equivalent, prefer the one that unblocks other work
-   - Run `ask info <id> 2>&1 | head -20` to preview the task — this caps the output at 20 lines so long descriptions do not flood the screen
+   - Run `do info <id> 2>&1 | head -20` to preview the task — this caps the output at 20 lines so long descriptions do not flood the screen
 
 3. **Mark the task started**:
-   - Run `ask start <id>`
+   - Run `do start <id>`
 
 4. **Delegate to a fresh sub-agent**:
    - Spawn a **new sub-agent** with a self-contained prompt that includes:
      - The task ID and a one-line summary of what the task is about
-     - Instruction to run `ask info <id>` as its **first action** to get the full description and all annotations (do not paste the description inline — the sub-agent fetches it fresh, keeping the prompt short)
+     - Instruction to run `do info <id>` as its **first action** to get the full description and all annotations (do not paste the description inline — the sub-agent fetches it fresh, keeping the prompt short)
      - The absolute path of the project root
-     - Instruction to run `ask annotate <id> "<progress notes>"` as it works
+     - Instruction to run `do annotate <id> "<progress notes>"` as it works
      - Instruction to commit all changes to git when done
      - Instruction to **not** mark the task done (the orchestrator does that)
    - The sub-agent must complete all implementation, tests, and a git commit before returning
    - Wait for the sub-agent to finish
 
 5. **Close and record**:
-   - Run `ask done <id>` to mark the task complete
-   - Run `ask annotate <id> "<summary of what was delivered>"` if the sub-agent did not already add a final annotation
+   - Run `do done <id>` to mark the task complete
+   - Run `do annotate <id> "<summary of what was delivered>"` if the sub-agent did not already add a final annotation
 
 6. **Auto-progress**:
    - Immediately return to step 1 and select the next pending task
diff --git a/prompts/skills/auditing-code-quality/SKILL.md b/prompts/skills/auditing-code-quality/SKILL.md
index 4e5f700..a0465ec 100644
--- a/prompts/skills/auditing-code-quality/SKILL.md
+++ b/prompts/skills/auditing-code-quality/SKILL.md
@@ -91,12 +91,12 @@ create a task for every HIGH and MEDIUM severity finding. Each task should:
 **Exact command format** — keep each part as a separate argument, never quoted together:
 
 ```bash
-ask add priority:H +code-quality "Refactor UserService to fix SRP violation"
-ask add priority:M +code-quality "Fix high cognitive complexity in parser.go"
+do add priority:H +code-quality "Refactor UserService to fix SRP violation"
+do add priority:M +code-quality "Fix high cognitive complexity in parser.go"
 ```
 
 Do NOT do this (causes tag to land in description):
 ```bash
-ask add "+code-quality Fix foo"          # wrong: tag+desc quoted as one arg
-ask add "+code-quality -p M Fix foo"     # wrong: everything in one quoted arg
+do add "+code-quality Fix foo"          # wrong: tag+desc quoted as one arg
+do add "+code-quality -p M Fix foo"     # wrong: everything in one quoted arg
 ```