fix(speckit-ext): address review on parallelization prose

- gather-context: make the fan-out a sub-bullet so the assembled plan body
  doesn't double-number step 2
- implement-exec: clarify the main agent journals each [P] task serially
  (never the subagents), keeping .spec-context.json writes foreground/race-free
- changelog: align the 0.10.0 date

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: alfredoperez

speckit-extension/CHANGELOG.md

@@ -6,7 +6,7 @@ All notable changes to the **spec-kit extension** (`id: companion`) are document
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/); this extension follows [Semantic Versioning](https://semver.org/).
 
-## [0.10.0] - 2026-06-15
+## [0.10.0] - 2026-06-16
 
 ### Added
 - **The pipeline runs as fast as your assistant allows.** When your AI assistant can work on several things at once, the Companion steps now spread the work out: it reads different parts of the codebase side by side while investigating, flags which tasks are independent enough to run together, and builds those independent tasks at the same time during implementation. Assistants that can't do that simply run each step the usual one-at-a-time way and produce the exact same result — nothing to turn on, nothing breaks.

speckit-extension/commands/speckit.companion.implement.md

@@ -29,7 +29,7 @@ If you cannot spawn subagents, do all of it sequentially — no error, identical
 
 2. Execute tasks in dependency order:
    - Complete each layer before the next: Setup → Foundational → Core → Integration → Polish.
-   - If you support subagents, run each `[P]` batch (different files, no incomplete dependency) concurrently — one subagent per task — and journal each as it finishes (timing rules unchanged). Same-file or dependent tasks stay ordered. No subagent support → run `[P]` tasks sequentially. A project may route task types to specialist subagents via a hook (the agent-routing seam).
+   - If you support subagents, run each `[P]` batch (different files, no incomplete dependency) concurrently — one subagent per task. As each finishes, the main agent (never the subagents) records it one at a time, so writes to `.spec-context.json` stay foreground and never race (timing rules unchanged). Same-file or dependent tasks stay ordered. No subagent support → run `[P]` tasks sequentially. A project may route task types to specialist subagents via a hook (the agent-routing seam).
    - Halt on a failed non-parallel task and report the cause; for `[P]` tasks, continue the others and report the failure.
 
 3. After completing a task, mark it `- [x]` in `tasks.md`.

speckit-extension/commands/speckit.companion.plan.md

@@ -25,8 +25,7 @@ If you cannot spawn subagents, do all of it sequentially — no error, identical
 <!-- /speckit-companion:part parallel -->
 
 1. Read `.specify/feature.json` for the feature directory; load `<feature_directory>/spec.md` and `.specify/memory/constitution.md` if present.
-
-2. If you support subagents, fan these reads out in parallel (one per area) and collect findings; otherwise read sequentially.
+   - If you support subagents, fan these reads out in parallel (one per area) and collect findings; otherwise read sequentially.
 
 2. Create `<feature_directory>/plan.md` with these sections, in order:
    - **Summary** — the primary requirement plus the technical approach in 2–4 sentences.

speckit-extension/nodes/implement/implement-exec.md

@@ -12,7 +12,7 @@ reads: []
 
 2. Execute tasks in dependency order:
    - Complete each layer before the next: Setup → Foundational → Core → Integration → Polish.
-   - If you support subagents, run each `[P]` batch (different files, no incomplete dependency) concurrently — one subagent per task — and journal each as it finishes (timing rules unchanged). Same-file or dependent tasks stay ordered. No subagent support → run `[P]` tasks sequentially. A project may route task types to specialist subagents via a hook (the agent-routing seam).
+   - If you support subagents, run each `[P]` batch (different files, no incomplete dependency) concurrently — one subagent per task. As each finishes, the main agent (never the subagents) records it one at a time, so writes to `.spec-context.json` stay foreground and never race (timing rules unchanged). Same-file or dependent tasks stay ordered. No subagent support → run `[P]` tasks sequentially. A project may route task types to specialist subagents via a hook (the agent-routing seam).
    - Halt on a failed non-parallel task and report the cause; for `[P]` tasks, continue the others and report the failure.
 
 3. After completing a task, mark it `- [x]` in `tasks.md`.

speckit-extension/nodes/plan/gather-context.md

@@ -5,6 +5,5 @@ command: plan
 reads: []
 ---
 1. Read `.specify/feature.json` for the feature directory; load `<feature_directory>/spec.md` and `.specify/memory/constitution.md` if present.
-
-2. If you support subagents, fan these reads out in parallel (one per area) and collect findings; otherwise read sequentially.
+   - If you support subagents, fan these reads out in parallel (one per area) and collect findings; otherwise read sequentially.
 

speckit-extension/tests/golden/commands/commands__speckit.companion.implement.md

@@ -29,7 +29,7 @@ If you cannot spawn subagents, do all of it sequentially — no error, identical
 
 2. Execute tasks in dependency order:
    - Complete each layer before the next: Setup → Foundational → Core → Integration → Polish.
-   - If you support subagents, run each `[P]` batch (different files, no incomplete dependency) concurrently — one subagent per task — and journal each as it finishes (timing rules unchanged). Same-file or dependent tasks stay ordered. No subagent support → run `[P]` tasks sequentially. A project may route task types to specialist subagents via a hook (the agent-routing seam).
+   - If you support subagents, run each `[P]` batch (different files, no incomplete dependency) concurrently — one subagent per task. As each finishes, the main agent (never the subagents) records it one at a time, so writes to `.spec-context.json` stay foreground and never race (timing rules unchanged). Same-file or dependent tasks stay ordered. No subagent support → run `[P]` tasks sequentially. A project may route task types to specialist subagents via a hook (the agent-routing seam).
    - Halt on a failed non-parallel task and report the cause; for `[P]` tasks, continue the others and report the failure.
 
 3. After completing a task, mark it `- [x]` in `tasks.md`.

speckit-extension/tests/golden/commands/commands__speckit.companion.plan.md

@@ -25,8 +25,7 @@ If you cannot spawn subagents, do all of it sequentially — no error, identical
 <!-- /speckit-companion:part parallel -->
 
 1. Read `.specify/feature.json` for the feature directory; load `<feature_directory>/spec.md` and `.specify/memory/constitution.md` if present.
-
-2. If you support subagents, fan these reads out in parallel (one per area) and collect findings; otherwise read sequentially.
+   - If you support subagents, fan these reads out in parallel (one per area) and collect findings; otherwise read sequentially.
 
 2. Create `<feature_directory>/plan.md` with these sections, in order:
    - **Summary** — the primary requirement plus the technical approach in 2–4 sentences.