Stronger wording to force skill invocation

Author: etr

skills/build-unplanned-feature/SKILL.md

@@ -21,7 +21,9 @@ Store the raw description for clarification.
 
 ### Step 2: Clarify Requirements
 
-Invoke: `Skill(skill="groundwork:understanding-feature-requests")`
+**You MUST call the Skill tool now:** `Skill(skill="groundwork:understanding-feature-requests")`
+
+Do NOT attempt to gather requirements yourself. The skill handles this.
 
 Follow the skill to gather:
 - Problem being solved
@@ -84,7 +86,10 @@ Then use `AskUserQuestion` to ask:
 
 ### Step 5: Execute Implementation
 
-Invoke: `Skill(skill="groundwork:implement-feature")`
+**CRITICAL INSTRUCTION: You MUST call the Skill tool now:**
+  `Skill(skill="groundwork:implement-feature")`
+
+Do NOT write implementation plans, do NOT create files, do NOT explore code, do NOT start coding. You are NOT permitted to implement the feature yourself. The implement-feature skill handles ALL implementation including worktree creation, TDD, validation, and merge. Call it NOW.
 
 Session context provides:
 - **identifier**: FEATURE-<slug>
@@ -93,13 +98,6 @@ Session context provides:
 - **action-items**: [from requirements]
 - **acceptance-criteria**: [from clarification]
 
-The `implement-feature` skill handles:
-1. Worktree creation via `use-git-worktree`
-2. TDD workflow for each requirement
-3. Multi-agent validation
-4. Worktree finalization (prompts user for merge decision)
-5. Completion reporting
-
 ---
 
 ## Reference

skills/execute-task/SKILL.md

@@ -141,20 +141,17 @@ Then use `AskUserQuestion` to ask:
 
 1. **Update status** - Change task to `**Status:** In Progress`
 
-2. **Invoke implementation** - `Skill(skill="groundwork:implement-feature")`
-
-Session context provides:
-- **identifier**: TASK-NNN
-- **title**: [Task Title]
-- **merge-mode**: `env`
-- **action-items**: [from task file]
-- **acceptance-criteria**: [from task file]
-
-The `implement-feature` skill handles:
-- Worktree creation via `use-git-worktree`
-- TDD implementation
-- Multi-agent validation
-- Worktree finalization (merge based on `GROUNDWORK_AUTO_MERGE` env var)
+2. **CRITICAL INSTRUCTION: You MUST call the Skill tool now:**
+     `Skill(skill="groundwork:implement-feature")`
+
+   Do NOT start implementing the task yourself. Do NOT create files, write code, or begin TDD directly. The implement-feature skill handles ALL of this. Call it NOW.
+
+   Session context provides:
+   - **identifier**: TASK-NNN
+   - **title**: [Task Title]
+   - **merge-mode**: `env`
+   - **action-items**: [from task file]
+   - **acceptance-criteria**: [from task file]
 
 ### Step 8: Complete Task
 

skills/implement-feature/SKILL.md

@@ -26,7 +26,9 @@ Check session context for worktree parameters:
 
 If no identifier create a new identifier yourself in the format FEATURE-slug. Otherwise, use the identifier provided.
 
-Invoke: `Skill(skill="groundwork:use-git-worktree", args="<identifier>")`
+**You MUST call the Skill tool now:** `Skill(skill="groundwork:use-git-worktree", args="<identifier>")`
+
+Do NOT create branches or worktrees with git commands directly. The skill handles setup and baseline validation.
 
 Wait for worktree creation to complete and baseline tests to pass.
 
@@ -50,7 +52,9 @@ Store the gathered context for use in subsequent steps.
 
 ### Step 3: Execute TDD
 
-Invoke: `Skill(skill="groundwork:test-driven-development")`
+**You MUST call the Skill tool now:** `Skill(skill="groundwork:test-driven-development")`
+
+Do NOT write implementation code without first loading the TDD skill. It enforces red-green-refactor discipline.
 
 For each action item:
 1. Write failing test
@@ -101,7 +105,9 @@ Output verification results:
 
 ### Step 6: Multi-Agent Verification
 
-This is mandatory. Invoke: `Skill(skill="groundwork:validation-loop")`
+**You MUST call the Skill tool now:** `Skill(skill="groundwork:validation-loop")`
+
+Do NOT declare implementation complete or skip validation. This step is non-negotiable.
 
 This runs verification agents in parallel with autonomous fix-and-retry.
 

skills/next-task/SKILL.md

@@ -60,10 +60,7 @@ Parse the tasks to find the next task to work on:
 
 ### Step 4: Delegate to Execute Task
 
-Once a task is identified, invoke the `groundwork:execute-task` skill with the task identifier (e.g., `TASK-004`).
+Once a task is identified, **you MUST call the Skill tool:**
+  `Skill(skill="groundwork:execute-task", args="TASK-NNN")`
 
-The execute-task skill will:
-1. Load full project context (PRD, architecture, tasks)
-2. Present the task summary
-3. Execute the task
-4. Complete the task and offer to continue
+Do NOT load project context, present summaries, or begin task execution yourself. The execute-task skill handles the complete workflow. Call it NOW with the identified task identifier.

skills/tasks/SKILL.md

@@ -285,12 +285,9 @@ Iterate until user approves, then write to `specs/tasks.md`.
 
 After user approves the task list and before writing final version:
 
-Invoke: `Skill(skill="groundwork:task-validation-loop")`
+**You MUST call the Skill tool now:** `Skill(skill="groundwork:task-validation-loop")`
 
-This runs verification agents to ensure:
-- All PRD requirements have corresponding tasks
-- Tasks are assigned to correct architecture components
-- UI tasks include design system and accessibility criteria
+Do NOT skip validation or declare the task list complete without running this. It ensures PRD coverage, architecture alignment, and design system compliance.
 
 **Wait for validation loop to pass before proceeding.**