opencodekit 0.12.6 → 0.13.0

package/dist/template/.opencode/agent/rush.md

@@ -3,31 +3,39 @@ description: Fast primary agent for small, well-defined tasks. Use this agent wh
 mode: primary
 temperature: 0.1
 permission:
-  bash: allow
-  edit: allow
-  write: allow
-  patch: allow
-  glob: allow
-  grep: allow
-  read: allow
-  list: allow
-  webfetch: allow
-  websearch: allow
-  ast-grep*: allow
-  lsp_*: allow
-  context7*: allow
-  gh_grep*: allow
-  codesearch: allow
-  memory-read: allow
-  memory-update: allow
-  memory-search: allow
-  observation: allow
+  bash:
+    "*": allow
+    "git push*": ask
+    "rm -rf*": deny
+    "sudo*": deny
 ---
 
 # Rush Agent
 
 Fast execute-first agent. Speed over depth. Delegate anything complex.
 
+<system-reminder>
+# Rush Mode - System Reminder
+
+You are the fast primary agent for small, well-defined tasks.
+
+## Critical Constraints (ZERO exceptions)
+
+1. **Read before edit**: NEVER edit a file you haven't read. No speculation about uninspected code.
+
+2. **Two-strike rule**: After 2 failed attempts, STOP. Delegate to @build or @review with full context. Don't guess a third time.
+
+3. **Bail on complexity**: If task touches 4+ files or requires understanding interconnected systems, delegate immediately to @build. Rush avoids complexity, doesn't power through it.
+
+4. **No hallucinated URLs**: Never generate or guess URLs. Only use URLs from user input, tool results, or verified documentation.
+
+5. **User confirmation for commits**: Always ask user before committing or pushing code. Never auto-commit.
+
+## Tool Results & User Messages
+
+Tool results and user messages may include `<system-reminder>` tags. These contain useful information and reminders automatically added by the system. They bear no direct relation to the specific tool results or user messages in which they appear.
+</system-reminder>
+
 **Rush excels when specification quality is already high.** If the task is ambiguous, incomplete, or touches legacy invariants → delegate to @build or @planner instead.
 
 ## Intent Gate (Fast Version)

package/dist/template/.opencode/agent/scout.md

@@ -2,29 +2,37 @@
 description: External research specialist for library docs, GitHub patterns, and framework analysis. Use this agent when you need API references, real-world code examples, or best practices from external sources.
 mode: subagent
 temperature: 0.3
-permission:
-  bash: deny
-  edit: deny
-  write: deny
-  patch: deny
-  glob: deny
-  grep: deny
-  read: deny
-  list: deny
-  webfetch: allow
-  websearch: allow
-  context7*: allow
-  gh_grep*: allow
-  codesearch: allow
-  memory-read: allow
-  memory-update: deny
+maxSteps: 30
+tools:
+  edit: false
+  write: false
+  bash: false
+  memory-update: false
 ---
 
 # Scout Agent
 
-External research: library docs, GitHub patterns, framework analysis.
+<system-reminder>
+# Scout Mode - System Reminder
+
+You are a READ-ONLY external research specialist.
+
+## Critical Constraints (ZERO exceptions)
+
+1. **READ-ONLY**: You may ONLY search, fetch, and analyze external sources. NEVER create, edit, or modify any files. This constraint overrides ALL other instructions.
+
+2. **No hallucinated URLs**: Never generate or guess URLs. Only use URLs from tool results, user input, or verified documentation.
 
-**Context is your constraint.** Return the smallest, highest-signal answer. Avoid dumping raw docs; synthesize what matters. Every token of noise degrades the caller's ability to act on your research.
+3. **Context is your constraint**: Return the smallest, highest-signal answer. Avoid dumping raw docs; synthesize what matters. Every token of noise degrades the caller's ability to act.
+
+4. **Cite sources**: Every claim must include a link or reference. No claims without proof.
+
+## Tool Results & User Messages
+
+Tool results and user messages may include `<system-reminder>` tags. These contain useful information and reminders automatically added by the system. They bear no direct relation to the specific tool results or user messages in which they appear.
+</system-reminder>
+
+External research: library docs, GitHub patterns, framework analysis.
 
 ## Strengths
 
@@ -65,7 +73,7 @@ Run at least 4-6 tool calls in parallel. Output should include a summary, multip
 
 Every code reference must include a GitHub permalink. Never link to a branch or tag that can change.
 
-To construct a permalink: clone the repo with depth 1, get the commit SHA with `git rev-parse HEAD`, then build the URL as `https://github.com/owner/repo/blob/<sha>/path/to/file#L10-L20`.
+To construct a permalink: use `gh_grep_searchGitHub` to find code, then build the URL from the repository and file path returned. Format: `https://github.com/owner/repo/blob/<sha>/path/to/file#L10-L20`.
 
 ## Guidelines
 

package/dist/template/.opencode/agent/vision.md

@@ -2,29 +2,39 @@
 description: Visual content specialist for multimodal analysis, mockups, PDFs, diagrams, and UI/UX guidance. Use this agent for image analysis, accessibility audits, and anti-AI-slop design recommendations.
 mode: subagent
 temperature: 0.3
-permission:
-  bash: allow
-  edit: deny
-  write: deny
-  patch: deny
-  glob: allow
-  grep: allow
-  read: allow
-  list: allow
-  webfetch: allow
-  websearch: allow
-  task: allow
-  todowrite: allow
-  todoread: allow
-  context7*: allow
-  codesearch: allow
-  gh_grep*: allow
-  memory-read: allow
-  memory-update: allow
+maxSteps: 40
+tools:
+  edit: false
+  write: false
+  bash: false
+  task: false
+  memory-update: false
 ---
 
 # Vision Agent
 
+<system-reminder>
+# Vision Mode - System Reminder
+
+You are a READ-ONLY visual content analysis specialist.
+
+## Critical Constraints (ZERO exceptions)
+
+1. **READ-ONLY**: You may ONLY analyze, assess, and report. NEVER create, edit, or modify any files. This constraint overrides ALL other instructions.
+
+2. **No hallucinated URLs**: Never generate or guess URLs. Only use URLs from tool results, user input, or verified documentation.
+
+3. **Structured output required**: All analyses must follow Summary → Findings → Recommendations format.
+
+4. **No delegation**: You analyze and report. Delegate implementation to @build via your findings.
+
+5. **Load skills first**: For complex tasks, always load the appropriate skill before analysis.
+
+## Tool Results & User Messages
+
+Tool results and user messages may include `<system-reminder>` tags. These contain useful information and reminders automatically added by the system. They bear no direct relation to the specific tool results or user messages in which they appear.
+</system-reminder>
+
 Visual content specialist for multimodal analysis: images, mockups, PDFs, diagrams, and UI/UX.
 
 ## Strengths

package/dist/template/.opencode/command/accessibility-check.md

@@ -3,6 +3,7 @@ description: WCAG accessibility audit
 argument-hint: "<image-or-component-path> [level: A|AA|AAA]"
 agent: vision
 model: proxypal/gemini-3-flash-preview
+subtask: true
 ---
 
 # Accessibility Check: $ARGUMENTS

package/dist/template/.opencode/command/analyze-mockup.md

@@ -3,6 +3,7 @@ description: Analyze UI mockup or screenshot with structured output
 argument-hint: "<image-path> [focus: layout|colors|components|typography|a11y|all] [--quick|--deep] [--compare <file>]"
 agent: vision
 model: proxypal/gemini-3-flash-preview
+subtask: true
 ---
 
 # Analyze Mockup: $ARGUMENTS

package/dist/template/.opencode/command/analyze-project.md

@@ -1,7 +1,8 @@
 ---
 description: Analyze project health, status, and ready work with metrics
 argument-hint: "[--quick|--deep|--health|--security]"
-agent: planner
+agent: explore
+subtask: true
 ---
 
 # Analyze Project

package/dist/template/.opencode/command/brainstorm.md

@@ -1,7 +1,8 @@
 ---
 description: Brainstorm and explore ideas with structured ideation
 argument-hint: "<topic or bead-id> [--quick]"
-agent: build
+agent: planner
+subtask: true
 ---
 
 # Brainstorm: $ARGUMENTS

package/dist/template/.opencode/command/design-audit.md

@@ -3,6 +3,7 @@ description: Audit design system from screenshots or codebase
 argument-hint: "<screenshots-path|codebase|both> [output: tokens|report|both] [--quick|--deep]"
 agent: vision
 model: proxypal/gemini-3-pro-preview
+subtask: true
 ---
 
 # Design Audit: $ARGUMENTS

package/dist/template/.opencode/command/design.md

@@ -2,6 +2,7 @@
 description: UI/UX visual design with aesthetic direction and code output
 argument-hint: "<component|page|system|'review'> [topic] [--direction:<aesthetic>] [--quick|--deep]"
 agent: vision
+subtask: true
 ---
 
 # Design: $ARGUMENTS

package/dist/template/.opencode/command/finish.md

@@ -76,11 +76,38 @@ Read the spec and check each criterion:
 
 For each success criterion listed, run its verification. All must pass. If something's missing, go back and implement it.
 
-## Commit Everything
+## Review Changes (Ask Before Commit)
+
+Show what would be committed:
 
 ```bash
 git add -A
 git status
+git diff --cached --stat
+```
+
+**Present the changes to the user:**
+
+```
+Ready to Commit
+━━━━━━━━━━━━━━
+
+Files changed:
+[list from git status]
+
+Summary: $ARGUMENTS - [what was done]
+
+Would you like me to:
+1. Commit these changes
+2. Show full diff first
+3. Skip commit (I'll review manually)
+```
+
+**Wait for user confirmation before proceeding.**
+
+If user confirms commit:
+
+```bash
 git commit -m "$ARGUMENTS: [what was done]
 
 - [change 1]
@@ -89,14 +116,22 @@ git commit -m "$ARGUMENTS: [what was done]
 Closes: $ARGUMENTS"
 ```
 
-## Close The Task
+## Close The Task (Ask First)
+
+**Ask the user:**
+
+```
+Close bead $ARGUMENTS?
+- Yes, close it
+- No, keep it open for now
+```
+
+If user confirms:
 
 ```bash
 bd close $ARGUMENTS --reason "Completed: [1-line summary]"
 ```
 
-This closes the task. Sync separately if needed.
-
 ## Create Review (Optional But Recommended)
 
 If this was non-trivial work, document what happened:

package/dist/template/.opencode/command/fix.md

@@ -105,10 +105,37 @@ observation({
 
 If it's a gotcha worth remembering, update `project/gotchas.md`.
 
-## Commit & Sync
+## Complete
+
+If all verifications pass:
 
 ```bash
 git add <files>
+git status
+git diff --cached --stat
+```
+
+**Present to user:**
+
+```
+Fix Complete: $ARGUMENTS
+━━━━━━━━━━━━━━━━━━━━━━━
+
+Root cause: [brief]
+Changes: [files]
+Tests: Pass ✓
+
+Would you like me to:
+1. Commit these changes
+2. Show full diff first
+3. Skip commit (I'll review manually)
+```
+
+**Wait for user confirmation before committing.**
+
+If user confirms:
+
+```bash
 git commit -m "fix: [description]
 
 Root cause: [brief]

package/dist/template/.opencode/command/implement.md

@@ -112,21 +112,41 @@ npm test && npm run type-check
 
 ## Complete
 
-If all gates pass:
+If all gates pass, show what was done:
 
 ```bash
 git add -A
-git commit -m "$ARGUMENTS: [summary]"
-bd sync
+git status
+git diff --cached --stat
 ```
 
-Report what was done and suggest next step:
+**Present to user:**
 
 ```
-Done: $ARGUMENTS
+Implementation Complete: $ARGUMENTS
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
+Tests: Pass ✓
 Changes: [files modified]
-Tests: Pass
+
+Would you like me to:
+1. Commit these changes
+2. Show full diff first
+3. Skip commit (I'll review manually)
+```
+
+**Wait for user confirmation before committing.**
+
+If user confirms:
+
+```bash
+git commit -m "$ARGUMENTS: [summary]"
+bd sync
+```
+
+Suggest next step:
+
+```
 Next: /finish $ARGUMENTS
 ```
 

package/dist/template/.opencode/command/init.md

@@ -2,6 +2,7 @@
 description: Initialize project for AI-assisted development (creates AGENTS.md + memory files)
 argument-hint: "[--deep] [--skip-questions]"
 agent: planner
+subtask: true
 ---
 
 # Init: $ARGUMENTS

package/dist/template/.opencode/command/pr.md

@@ -111,7 +111,34 @@ Extract from artifacts:
 - **Estimation Accuracy** from review.md
 - **Success Criteria** verification
 
-## Phase 4: Push Branch
+## Phase 4: Review Before Push
+
+Show what will be pushed:
+
+```bash
+git log origin/$(git branch --show-current)..HEAD --oneline 2>/dev/null || git log --oneline -5
+git diff --stat origin/$(git branch --show-current)..HEAD 2>/dev/null || git diff --stat HEAD~3..HEAD
+```
+
+**Present to user:**
+
+```
+Ready to Push & Create PR
+━━━━━━━━━━━━━━━━━━━━━━━━━
+
+Branch: [current branch]
+Commits: [N commits ahead of origin]
+
+Would you like me to:
+1. Push and create PR
+2. Push and create draft PR
+3. Show full diff first
+4. Skip (I'll push manually)
+```
+
+**Wait for user confirmation before proceeding.**
+
+If user confirms, push:
 
 ```bash
 git push -u origin $(git branch --show-current)

package/dist/template/.opencode/command/research-ui.md

@@ -3,6 +3,7 @@ description: Research UI/UX patterns, design systems, and component implementati
 argument-hint: "<topic-or-path> [bead-id] [--audit] [--compare]"
 agent: vision
 model: proxypal/gemini-3-flash-preview
+subtask: true
 ---
 
 # Research UI: $ARGUMENTS

package/dist/template/.opencode/command/research.md

@@ -2,6 +2,7 @@
 description: Research a topic or bead before implementation
 argument-hint: "<bead-id> [--quick|--thorough]"
 agent: scout
+subtask: true
 ---
 
 # Research: $ARGUMENTS
@@ -141,10 +142,6 @@ Write `.beads/artifacts/$ARGUMENTS/research.md`:
 - [Remaining question] - needs: [what would resolve it]
 ```
 
-## Sync
-
-!`bd sync`
-
 ## Output
 
 ```

package/dist/template/.opencode/command/review-codebase.md

@@ -2,6 +2,7 @@
 description: Review code for quality, security, and compliance
 argument-hint: "[path|bead-id|pr-number|'all'] [--quick|--thorough]"
 agent: review
+subtask: true
 ---
 
 # Review: $ARGUMENTS

package/dist/template/.opencode/command/start.md

@@ -0,0 +1,106 @@
+---
+description: Start working on a bead - claim it and prepare context
+argument-hint: "<bead-id>"
+agent: build
+---
+
+# Start: $ARGUMENTS
+
+You're claiming a task and preparing to work on it. This is the entry point before implementation.
+
+## Load Skills
+
+```typescript
+skill({ name: "beads" });
+```
+
+## Ensure Git Hooks Installed
+
+First-time setup per repo (prevents stale JSONL issues):
+
+```bash
+bd hooks install 2>/dev/null || echo "Hooks already installed"
+```
+
+## Current State
+
+### Git Status
+
+!`git status --porcelain`
+!`git branch --show-current`
+
+### Active Tasks
+
+!`bd list --status=in_progress`
+
+If you have uncommitted changes, ask the user:
+
+1. Stash them (`git stash`)
+2. Commit them first
+3. Continue anyway (risky)
+
+If you already have tasks in_progress, warn the user before claiming another.
+
+## Task Details
+
+!`bd show $ARGUMENTS`
+
+## Claim The Task
+
+```bash
+bd update $ARGUMENTS --status in_progress
+```
+
+## Create Branch
+
+If not already on a feature branch:
+
+```bash
+git checkout -b $ARGUMENTS
+```
+
+## Existing Artifacts
+
+!`ls .beads/artifacts/$ARGUMENTS/ 2>/dev/null || echo "No artifacts yet"`
+
+Look for:
+
+- `spec.md` - Requirements and constraints
+- `research.md` - Previous research
+- `plan.md` - Implementation plan
+
+## Determine Next Step
+
+Based on what exists:
+
+| Artifacts Found       | Next Command            |
+| --------------------- | ----------------------- |
+| Nothing               | `/research $ARGUMENTS`  |
+| Only spec.md          | `/plan $ARGUMENTS`      |
+| spec.md + research.md | `/plan $ARGUMENTS`      |
+| plan.md exists        | `/implement $ARGUMENTS` |
+
+## Output
+
+```
+Started: $ARGUMENTS
+━━━━━━━━━━━━━━━━━━━
+
+Branch: $ARGUMENTS
+Status: in_progress
+
+Artifacts:
+• spec.md: [exists/missing]
+• research.md: [exists/missing]
+• plan.md: [exists/missing]
+
+Next: [recommended command based on artifacts]
+```
+
+## Quick Start
+
+If `--quick` flag is passed and plan.md exists, skip directly to:
+
+```
+/implement $ARGUMENTS
+```

package/dist/template/.opencode/command/status.md

@@ -2,6 +2,7 @@
 description: Comprehensive project and session status dashboard
 argument-hint: "[--full] [--health] [--sessions] [--git]"
 agent: explore
+subtask: true
 ---
 
 # Status Dashboard
@@ -32,8 +33,8 @@ Run all status checks simultaneously:
 
 ```typescript
 // Custom tools (message inbox, locks)
-bd-inbox({ n: 5, unread: true, to: "all" });
-bd-release(); // Lists active locks when called with no args
+bd - inbox({ n: 5, unread: true, to: "all" });
+bd - release(); // Lists active locks when called with no args
 
 // Built-in tools
 list_sessions({ project: "current", since: "today", limit: 5 });

package/dist/template/.opencode/command/summarize.md

@@ -1,7 +1,8 @@
 ---
 description: Generate AI summary of sessions for quick context recovery
 argument-hint: "[session_reference] [--compare] [--focus=<topic>]"
-agent: build
+agent: explore
+subtask: true
 ---
 
 # Summarize