feat(skills): Enforce worktree/branch workflow in parallel_beads

home/roles/development/skills/beads_implement.md

@@ -54,8 +54,6 @@ When this command is invoked:
    - Read `thoughts/beads-{bead-id}/plan.md` FULLY
    - Check for any existing checkmarks (- [x]) indicating partial progress
    - Read any research at `thoughts/beads-{bead-id}/research.md`
-   - If plan's Success Criteria references contribution guidelines (e.g., "Per CONTRIBUTING.md:"),
-     verify the original CONTRIBUTING.md still exists and requirements are current
 
 5. **Mark bead in progress** (if not already):
    ```bash
@@ -129,10 +127,6 @@ All phases completed and automated verification passed:
 - {List manual verification items from plan}
 
 Let me know when manual testing is complete so I can close the bead.
-
-**Contribution guidelines compliance:**
-- {List any contribution guideline requirements that were part of Success Criteria}
-- {Note if any requirements could not be automated and need manual review}
 ```
 
 **STOP HERE and wait for user confirmation.**

home/roles/development/skills/beads_plan.md

@@ -51,32 +51,13 @@ When this command is invoked:
    - Any linked tickets or docs
    - Use Read tool WITHOUT limit/offset
 
-2. **Check for contribution guidelines**:
-
-   ```bash
-   # Check standard locations for contribution guidelines
-   for f in CONTRIBUTING.md .github/CONTRIBUTING.md docs/CONTRIBUTING.md; do
-     if [ -f "$f" ]; then
-       echo "Found: $f"
-       break
-     fi
-   done
-   ```
-
-   If found:
-   - Read the file fully
-   - Extract actionable requirements (testing, code style, documentation, PR conventions)
-   - These requirements MUST be incorporated into the plan's Success Criteria
-
-   If not found, note "No contribution guidelines found" and proceed.
-
-3. **Spawn initial research tasks**:
+2. **Spawn initial research tasks**:
    - **codebase-locator**: Find all files related to the task
    - **codebase-analyzer**: Understand current implementation
    - **codebase-pattern-finder**: Find similar features to model after
    - **thoughts-locator**: Find any existing plans or decisions
 
-4. **Read all files identified by research**:
+3. **Read all files identified by research**:
    - Read them FULLY into main context
    - Cross-reference with requirements
 
@@ -292,12 +273,6 @@ Always separate into two categories:
 - Performance under real conditions
 - Edge cases hard to automate
 
-**From Contribution Guidelines** (if CONTRIBUTING.md exists):
-- Include any testing requirements specified in guidelines
-- Include any code style/linting requirements
-- Include any documentation requirements
-- Reference the guideline: "Per CONTRIBUTING.md: {requirement}"
-
 ## Example Invocation
 
 ```

home/roles/development/skills/beads_research.md

@@ -51,18 +51,6 @@ When this command is invoked:
 - Use the Read tool WITHOUT limit/offset parameters
 - Read these files yourself in the main context before spawning sub-tasks
 
-### Step 1.5: Check for contribution guidelines
-
-Before spawning sub-agents, check if the repository has contribution guidelines:
-
-```bash
-for f in CONTRIBUTING.md .github/CONTRIBUTING.md docs/CONTRIBUTING.md; do
-  if [ -f "$f" ]; then echo "Found: $f"; break; fi
-done
-```
-
-If found, read the file and note key requirements. These should be included in the research document under a "## Contribution Guidelines" section if relevant to the research question.
-
 ### Step 2: Analyze and decompose the research question
 - Break down the query into composable research areas
 - Identify specific components, patterns, or concepts to investigate
@@ -155,10 +143,6 @@ status: complete
 ## Architecture Documentation
 {Current patterns, conventions found in codebase}
 
-## Contribution Guidelines
-{If CONTRIBUTING.md exists, summarize key requirements relevant to the research topic}
-{If no guidelines found, omit this section}
-
 ## Historical Context (from thoughts/)
 {Relevant insights from thoughts/ with references}
 

home/roles/development/skills/parallel_beads.md

@@ -42,7 +42,46 @@ AskUserQuestion with:
 - options from filtered bd ready output
 ```
 
-## Phase 2: Parallel Implementation
+## Phase 2: Worktree Setup
+
+Before launching implementation subagents, create worktrees for all selected beads:
+
+1. **Get repository name**:
+   ```bash
+   REPO_NAME=$(git remote get-url origin | sed 's|.*/||' | sed 's/\.git$//')
+   ```
+
+2. **For each selected bead**, create its worktree:
+   ```bash
+   BEAD_ID="[bead-id]"
+   # Check if worktree already exists
+   if [ -d "$HOME/wt/${REPO_NAME}/${BEAD_ID}" ]; then
+     echo "Worktree already exists: ~/wt/${REPO_NAME}/${BEAD_ID}"
+     # Ask user: remove and recreate, or skip this bead?
+   else
+     git worktree add -b "bead/${BEAD_ID}" "$HOME/wt/${REPO_NAME}/${BEAD_ID}"
+   fi
+   ```
+
+3. **Track created worktrees**:
+   Maintain a list of (bead_id, worktree_path) pairs for use in subagent instructions.
+
+4. **Report status**:
+   ```
+   Created worktrees:
+   - nixos-configs-abc → ~/wt/nixos-configs/nixos-configs-abc (branch: bead/nixos-configs-abc)
+   - nixos-configs-xyz → ~/wt/nixos-configs/nixos-configs-xyz (branch: bead/nixos-configs-xyz)
+
+   Skipped (existing worktree):
+   - nixos-configs-123 → Ask user for resolution
+   ```
+
+**Note**: If a worktree or branch already exists, ask the user before proceeding:
+- Remove existing worktree and branch, then recreate
+- Skip this bead
+- Use existing worktree as-is (risky - branch may have diverged)
+
+## Phase 3: Parallel Implementation
 
 For each selected bead, launch a subagent using the Task tool. All subagents should be launched in parallel (single message with multiple Task tool calls).
 
@@ -53,35 +92,60 @@ Each implementation subagent should receive these instructions:
 ```
 Work on bead [BEAD_ID]: [BEAD_TITLE]
 
-1. **Create worktree**:
-   - Branch name: `bead/[BEAD_ID]`
-   - Worktree path: `~/wt/[REPO_NAME]/[BEAD_ID]`
-   - Command: `git worktree add -b bead/[BEAD_ID] ~/wt/[REPO_NAME]/[BEAD_ID]`
+Worktree path: [WORKTREE_PATH]
 
-2. **Review the bead requirements**:
+## CRITICAL: Branch Verification (MUST DO FIRST)
+
+1. **Navigate to worktree**:
+   ```bash
+   cd [WORKTREE_PATH]
+   ```
+
+2. **Verify branch** (MANDATORY before ANY modifications):
+   ```bash
+   CURRENT_BRANCH=$(git branch --show-current)
+   echo "Current branch: $CURRENT_BRANCH"
+   pwd
+   ```
+
+   **ABORT CONDITIONS** - If ANY of these are true, STOP IMMEDIATELY:
+   - Branch is `main` or `master`
+   - Branch does not match `bead/[BEAD_ID]`
+
+   If you detect any abort condition:
+   ```
+   ABORTING: Branch verification failed.
+   Expected branch: bead/[BEAD_ID]
+   Actual branch: [CURRENT_BRANCH]
+   Working directory: [pwd output]
+
+   DO NOT PROCEED. Report this error to the orchestrator.
+   ```
+
+## After Verification Passes
+
+3. **Review the bead requirements**:
    - Run `bd show [BEAD_ID]` to understand the acceptance criteria
    - Note any external issue references (GitHub issues, Linear tickets, etc.)
 
-3. **Extract validation criteria**:
+4. **Extract validation criteria**:
    - Check for a plan: `thoughts/beads-[BEAD_ID]/plan.md`
    - If plan exists:
      - Read the plan and find the "Automated Verification" section
      - Extract each verification command (lines starting with `- [ ]` followed by a command)
      - Example: `- [ ] Tests pass: \`make test\`` → extract `make test`
-     - Note any "Per CONTRIBUTING.md:" requirements for additional validation
    - If no plan exists, use best-effort validation:
      - Check if `Makefile` exists → try `make test` and `make lint`
      - Check if `flake.nix` exists → try `nix flake check`
      - Check if `package.json` exists → try `npm test`
-     - **Check for CONTRIBUTING.md** → read and extract testing/linting requirements
      - If none found, note "No validation criteria found"
 
-4. **Implement the changes**:
+5. **Implement the changes**:
    - Work in the worktree directory
    - Complete all acceptance criteria listed in the bead
 
    After implementation, run validation:
-   - Execute each validation command from step 3
+   - Execute each validation command from step 4
    - Track results in this format:
      ```
      VALIDATION_RESULTS:
@@ -93,12 +157,12 @@ Work on bead [BEAD_ID]: [BEAD_TITLE]
      - Continue with PR creation (don't block)
      - Document failures in bead notes: `bd update [BEAD_ID] --notes="Validation failures: [list]"`
 
-5. **Commit and push**:
+6. **Commit and push**:
    - Stage all changes: `git add -A`
    - Create a descriptive commit message
    - Push the branch: `git push -u origin bead/[BEAD_ID]`
 
-6. **Create a PR**:
+7. **Create a PR**:
    - Detect hosting provider from origin URL: `git remote get-url origin`
    - If URL contains `github.com`, use `gh`; otherwise use `tea` (Gitea/Forgejo)
    - PR title: "[BEAD_ID] [BEAD_TITLE]"
@@ -158,13 +222,13 @@ Work on bead [BEAD_ID]: [BEAD_TITLE]
      | nix flake check | SKIP | command not found |"
      ```
 
-7. **Update bead status**:
+8. **Update bead status**:
    - Mark the bead as "in_review": `bd update [BEAD_ID] --status=in_review`
    - Add the PR URL to the bead notes: `bd update [BEAD_ID] --notes="$(bd show [BEAD_ID] --json | jq -r '.notes')
 
 PR: [PR_URL]"`
 
-8. **Report results**:
+9. **Report results**:
    - Return:
      - PR URL
      - Bead ID
@@ -177,15 +241,24 @@ PR: [PR_URL]"`
 
 ### Launching Subagents
 
+For each bead, substitute into the template:
+- `[BEAD_ID]` - the bead ID
+- `[BEAD_TITLE]` - the bead title
+- `[WORKTREE_PATH]` - the worktree path created in Phase 2
+
 Use `subagent_type: "general-purpose"` for implementation subagents. Launch all selected beads' subagents in a single message for parallel execution:
 
 ```
 <Task calls for each selected bead - all in one message>
 ```
 
+**Important**: The worktree paths were created in Phase 2. Use the exact paths that were created, e.g.:
+- `~/wt/nixos-configs/nixos-configs-abc`
+- `~/wt/nixos-configs/nixos-configs-xyz`
+
 Collect results from all subagents before proceeding.
 
-## Phase 3: Parallel Review
+## Phase 4: Parallel Review
 
 After all implementation subagents complete, launch review subagents for each PR.
 
@@ -220,7 +293,7 @@ Review PR for bead [BEAD_ID]
 
 Launch all review subagents in parallel.
 
-## Phase 4: Cleanup and Summary
+## Phase 5: Cleanup and Summary
 
 After reviews complete:
 
@@ -266,9 +339,21 @@ Example output:
 
 ## Error Handling
 
+- **Worktree creation failures** (Phase 2):
+  - If `git worktree add` fails (branch exists, path exists), prompt user:
+    - Remove existing and retry
+    - Skip this bead
+    - Use existing (with warning about potential divergence)
+  - Do NOT proceed to subagent launch until worktree is confirmed
+
+- **Branch verification failures** (subagent reports):
+  - If subagent reports it's on `main` or `master`, do NOT retry
+  - Mark bead as failed with reason "Branch verification failed"
+  - Continue with other beads but flag this as a critical issue
+  - Investigation required: the worktree may have been corrupted or not created properly
+
 - **Subagent failures**: If a subagent fails or times out, note it in the summary but continue with other beads
 - **PR creation failures**: Report the error but continue with reviews of successful PRs
-- **Worktree conflicts**: If a worktree already exists, ask the user if they want to remove it or skip that bead
 
 ## Resource Limits