Clarify captain-led director workflow

docs/modernization/director-workflow.md

@@ -24,15 +24,24 @@ The director is the main agent in the user thread. The director owns:
 
 - choosing the task group from `docs/modernization/tasks.md`
 - deciding whether delegation is worth the coordination cost
-- spawning one `gpt-5.4` captain per independent task group
+- spawning one `gpt-5.4` captain per independent task group, including
+  multiple teams when scopes are disjoint
 - integrating returned changes
 - running final validation
 - updating docs/debt/tasks
 - committing and pushing the verified slice
 - resetting conversation context after the slice when practical
 
-The director should not delegate the immediate blocking step. If the next local
+The director should keep local work minimal. It should not implement production
-action is required before anything else can proceed, do that locally.
+changes unless a required integration fix is smaller than another delegation
+round. Prefer giving captains ownership of implementation, focused validation,
+and worker coordination, then use the director context only for scope selection,
+conflict review, final validation, docs/task updates, commits, and pushes.
+
+The director may do a quick local blocking check before delegation, such as
+reading task rows, checking git status, or confirming that two scopes do not
+overlap. If the next action is substantive code or test work, delegate it to a
+captain or worker whenever the scope can be made clear.
 
 ### Team Captain
 
@@ -46,10 +55,13 @@ example:
 - dependency cleanup
 - test/hardening work
 
-The captain turns one task-group objective into smaller disjoint subtasks. If
+The captain owns implementation for its assigned task group. The captain turns
-nested subagents are available, the captain may spawn workers. If nested
+one task-group objective into smaller disjoint subtasks, spawns or requests
-subagents are not available in the current surface, the captain returns a worker
+workers when useful, reviews their changes, runs focused validation when cheap,
-plan and the director performs the second-level spawns.
+and returns an integration-ready result. If nested subagents are available, the
+captain may spawn workers. If nested subagents are not available in the current
+surface, the captain returns a worker plan and the director performs the
+second-level spawns without taking over implementation.
 
 The captain must not edit broad shared files unless assigned. The captain must
 not rewrite the task tracker or roadmap except for a requested status note.
@@ -96,25 +108,23 @@ has requested `gpt-5.4` captains, so use that override for captains.
 - Ask for compact final reports: changed files, result, validation, blockers,
   next recommendation.
 - Close completed agents after their results are integrated or rejected.
-- Prefer one captain and two or three workers over many agents. Coordination
+- Prefer the smallest number of teams that keeps disjoint work moving. Multiple
-  cost can exceed token savings.
+  captains are appropriate when task rows have non-overlapping write scopes and
+  can validate independently. Avoid many agents in one file family.
 
 ## Delegation Flow
 
-1. Director picks one `Ready` task or one small group of adjacent tasks from
+1. Director picks one or more `Ready` tasks from
-   `docs/modernization/tasks.md`.
+   `docs/modernization/tasks.md` with disjoint write scopes.
-2. Director identifies the immediate local blocking step and keeps it local.
+2. Director assigns each independent task group to a `gpt-5.4` captain.
-3. Director spawns one `gpt-5.4` captain for sidecar decomposition or an
+3. Captains implement directly or coordinate workers/explorers for disjoint
-   independent task group.
+   subtasks.
-4. Captain returns either:
+4. Captains run focused validation when cheap and return changed files,
-   - a worker plan with disjoint write scopes, or
+   validation, blockers, and integration notes.
-   - completed bounded edits if the captain can finish the task directly.
+5. Director reviews for scope conflicts and integrates returned changes.
-5. Director or captain spawns worker/explorer agents only for disjoint subtasks.
+6. Director runs the listed validation command or the quiet checkpoint wrapper.
-6. Workers edit only their assigned files and run focused validation when cheap.
+7. Director updates `tasks.md`, `debt.md`, and `roadmap.md` if tasks moved.
-7. Director reviews and integrates changes.
+8. Director commits and pushes the verified slice.
-8. Director runs the task's listed validation command.
-9. Director updates `tasks.md`, `debt.md`, and `roadmap.md` if the task moved.
-10. Director commits and pushes the verified slice.
 
 ## Director Prompt Template For A Captain
 
@@ -128,16 +138,18 @@ Validation: <COMMANDS>.
 
 Goal: <ONE PARAGRAPH>.
 
-First, split this into disjoint worker tasks suitable for gpt-5.4-mini or
+Own this task group through implementation. Split into disjoint worker tasks
-gpt-5.3-codex-spark. Use gpt-5.4 only for risky C++ behavior changes. Keep
+only when that is faster or safer than doing the bounded edits yourself. Use
-tasks small enough to validate. Do not edit outside the assigned scope. Other
+gpt-5.4 only for risky C++ behavior changes. Keep tasks small enough to
-agents may be working in parallel; do not revert unrelated changes.
+validate. Do not edit outside the assigned scope. Other agents may be working
+in parallel; do not revert unrelated changes.
 
 Return:
 - recommended worker tasks with model choice and file scope
-- any task you can safely complete yourself
+- completed edits or the exact blocker preventing them
+- focused validation run and result
 - risks/blockers
-- minimal context each worker needs
+- integration notes for the director
 ```
 
 ## Captain Prompt Template For A Worker