@ultrakit/ultrakit 0.0.0 → 0.0.2

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

@@ -0,0 +1,368 @@
+---
+description: Deep project initialization — generates hierarchical AGENTS.md knowledge bases across directories
+argument-hint: "[--create-new] [--max-depth=N] [--scope=DIR]"
+agent: build
+---
+
+# Init-Deep: $ARGUMENTS
+
+Deep project initialization. Generates hierarchical AGENTS.md files across the codebase for maximum agent context quality. Extends `/init` with directory-level knowledge bases.
+
+> **Prerequisite:** Run `/init` first for core AGENTS.md + tech-stack setup
+> **Related:** `/init-context` for planning workflow, `/init-user` for personalization
+
+## Options
+
+| Argument        | Default | Description                                  |
+| --------------- | ------- | -------------------------------------------- |
+| `--create-new`  | false   | Delete all existing AGENTS.md and regenerate |
+| `--max-depth=N` | 5       | Maximum directory depth to traverse          |
+| `--scope=DIR`   | .       | Limit generation to a specific directory     |
+
+## Load Skills
+
+```typescript
+skill({ name: "index-knowledge" });
+skill({ name: "memory-system" });
+skill({ name: "verification-before-completion" });
+```
+
+## Parse Arguments
+
+```typescript
+const args = {
+  createNew: $ARGUMENTS.includes("--create-new"),
+  maxDepth: parseInt($ARGUMENTS.match(/--max-depth=(\d+)/)?.[1] || "5", 10),
+  scope: $ARGUMENTS.match(/--scope=([^\s]+)/)?.[1] || ".",
+};
+```
+
+## Phase 1: Pre-Flight Check
+
+### 1.1 Verify Core Initialization
+
+```typescript
+// Check if /init has been run
+const agentsMd = tilth_tilth_files({ pattern: "AGENTS.md", scope: args.scope });
+const techStack = tilth_tilth_files({
+  pattern: "tech-stack.md",
+  scope: ".opencode/memory/project",
+});
+```
+
+**If no root AGENTS.md exists:**
+
+```
+⚠️  No root AGENTS.md found. Run /init first to set up core project knowledge.
+
+Options:
+1. Run /init now (recommended)
+2. Continue anyway (agents will have reduced context)
+3. Cancel
+```
+
+Wait for user selection.
+
+### 1.2 Check Memory for Previous Runs
+
+```typescript
+memory-search({ query: "AGENTS.md init-deep index-knowledge", limit: 5 });
+memory-search({ query: "project structure conventions", type: "observations", limit: 3 });
+```
+
+If previous init-deep was run recently (within 7 days), show:
+
+```
+Previous init-deep detected ([date]).
+Options:
+1. Update mode (preserve existing, fill gaps)
+2. Full regeneration (--create-new)
+3. Cancel
+```
+
+### 1.3 Preserve Existing AGENTS.md
+
+**If NOT `--create-new`:**
+
+```typescript
+// Read all existing AGENTS.md files
+const existingFiles = [];
+for (const file of agentsMd) {
+  const content = Read({ filePath: file, limit: 20 });
+  existingFiles.push({ path: file, firstLines: content });
+}
+// Store for reference during generation
+```
+
+**If `--create-new`:**
+
+```typescript
+// Read all existing for context extraction before deletion
+for (const file of agentsMd) {
+  const content = Read({ filePath: file });
+  // Extract key insights, conventions, anti-patterns
+  // These will be re-incorporated into regenerated files
+}
+```
+
+## Phase 2: Discovery + Analysis
+
+Delegate to the `index-knowledge` skill's Phase 1:
+
+```typescript
+skill({ name: "index-knowledge" });
+```
+
+The index-knowledge skill will:
+
+1. Launch parallel explore agents (project structure, entry points, conventions, anti-patterns, build/CI, test patterns)
+2. Run bash structural analysis (file counts, directory depth, code concentration)
+3. Read existing AGENTS.md files
+4. Merge all findings
+
+**Dynamic agent spawning** based on project scale:
+
+| Factor                   | Threshold | Additional Agents          |
+| ------------------------ | --------- | -------------------------- |
+| Total files              | >100      | +1 per 100 files           |
+| Total lines              | >10k      | +1 per 10k lines           |
+| Directory depth          | ≥4        | +2 for deep exploration    |
+| Large files (>500 lines) | >10 files | +1 for complexity hotspots |
+| Monorepo                 | detected  | +1 per package/workspace   |
+
+## Phase 3: Scoring + Location Decision
+
+Follow the index-knowledge scoring matrix:
+
+| Factor               | Weight | High Threshold | Source  |
+| -------------------- | ------ | -------------- | ------- |
+| File count           | 3x     | >20            | bash    |
+| Subdir count         | 2x     | >5             | bash    |
+| Code ratio           | 2x     | >70%           | bash    |
+| Unique patterns      | 1x     | Own config     | explore |
+| Module boundary      | 2x     | Has index.ts   | bash    |
+| Symbol density       | 2x     | >30 symbols    | LSP     |
+| Export count         | 2x     | >10 exports    | LSP     |
+| Reference centrality | 3x     | >20 refs       | LSP     |
+
+**Decision rules:**
+
+| Score    | Action                    |
+| -------- | ------------------------- |
+| Root (.) | ALWAYS create             |
+| >15      | Create AGENTS.md          |
+| 8-15     | Create if distinct domain |
+| <8       | Skip (parent covers)      |
+
+**Respect `--max-depth`**: Don't create AGENTS.md deeper than the specified depth.
+
+**Respect `--scope`**: Only analyze directories under the specified scope.
+
+## Phase 4: Generate AGENTS.md Hierarchy
+
+### 4.1 Root AGENTS.md
+
+**If root AGENTS.md already exists and NOT `--create-new`:**
+
+- Improve existing content using discovery findings
+- Fill gaps (missing sections, outdated info)
+- Preserve user-written conventions and gotchas
+- Add newly discovered patterns
+
+**If `--create-new` or no root AGENTS.md exists:**
+
+Generate root AGENTS.md following this structure (target <60 lines, max 150):
+
+```markdown
+# PROJECT KNOWLEDGE BASE
+
+**Generated:** {DATE}
+**Source:** /init-deep command
+
+## OVERVIEW
+
+{1-2 sentences: what this project does + core stack}
+
+## STRUCTURE
+
+{root}/
+├── {dir}/ # {non-obvious purpose only}
+└── {entry} # {entry point}
+
+## WHERE TO LOOK
+
+| Task   | Location | Notes |
+| ------ | -------- | ----- |
+| {task} | {path}   | {tip} |
+
+## COMMANDS
+
+{dev/test/build/lint commands — validated}
+
+## CONVENTIONS
+
+{ONLY deviations from standard patterns}
+
+## ANTI-PATTERNS
+
+{Explicitly forbidden patterns for THIS project}
+
+## GOTCHAS
+
+{Project-specific footguns}
+
+## NOTES
+
+{Important things not covered above}
+```
+
+**Key rules:**
+
+- Examples > explanations
+- Pointers > copies
+- Telegraphic style — no filler words
+- Never repeat what the parent AGENTS.md says
+- Never include generic advice that applies to all projects
+
+### 4.2 Subdirectory AGENTS.md (Parallel)
+
+Launch general agents for each scored location in ONE message:
+
+```typescript
+// All Task calls in single message = parallel execution
+for (const location of AGENTS_LOCATIONS) {
+  if (location.path === ".") continue; // Root already done
+
+  task({
+    subagent_type: "general",
+    description: `AGENTS.md for ${location.path}`,
+    prompt: `Generate AGENTS.md for: ${location.path}
+
+Context:
+- Reason: ${location.reason} (score: ${location.score})
+- This is a subdirectory AGENTS.md — 30-80 lines max
+- NEVER repeat parent AGENTS.md content
+- Sections: OVERVIEW (1 line), STRUCTURE (if >5 subdirs), WHERE TO LOOK, CONVENTIONS (only if different from parent), ANTI-PATTERNS
+- Telegraphic style — no filler words, no generic advice
+- Include a "See parent" reference: "For project-wide context, see ../AGENTS.md"
+
+Existing findings to incorporate:
+${location.findings}
+
+Write directly to ${location.path}/AGENTS.md`,
+  });
+}
+```
+
+### 4.3 Handle `--create-new` Cleanup
+
+If `--create-new` was specified:
+
+1. All old AGENTS.md files were read in Phase 1.3
+2. Key insights extracted from old files are re-incorporated
+3. New files replace old ones completely
+4. Report which files were regenerated
+
+## Phase 5: Review + Deduplicate
+
+For each generated file:
+
+1. **Remove generic advice** — anything that applies to all projects (e.g., "use consistent naming")
+2. **Remove parent duplicates** — child AGENTS.md should not repeat parent content
+3. **Trim to size limits** — root <150 lines, subdirectories <80 lines
+4. **Verify telegraphic style** — no filler words, no "Remember to..." phrasing
+5. **Check cross-references** — each child should reference its parent
+6. **Validate commands** — run `--version` or `--help` on detected commands where possible
+
+## Phase 6: Update Context README
+
+Update `.opencode/context/README.md` to reflect the new knowledge hierarchy:
+
+```typescript
+// Read existing context README
+const contextReadme = Read({ filePath: ".opencode/context/README.md" });
+
+// Append knowledge base section if not present
+// Include paths to all generated AGENTS.md files
+```
+
+## Phase 7: Record + Verify
+
+### 7.1 Record Findings
+
+```typescript
+observation({
+  type: "pattern",
+  title: "init-deep: hierarchical AGENTS.md generated",
+  narrative: `Generated ${count} AGENTS.md files across directories. Key locations: ${locations.join(", ")}. Mode: ${args.createNew ? "create-new" : "update"}.`,
+  concepts: "AGENTS.md, knowledge-base, init-deep, hierarchical",
+  confidence: "high",
+  files_modified: agentsMdPaths.join(", "),
+});
+```
+
+### 7.2 Verify
+
+```typescript
+// verification-before-completion skill already loaded
+// Check:
+// - [ ] Root AGENTS.md exists and is <150 lines
+// - [ ] Each subdirectory AGENTS.md exists and is <80 lines
+// - [ ] No child AGENTS.md repeats parent content
+// - [ ] All generated files are valid markdown
+// - [ ] Commands section lists validated commands
+// - [ ] Context README updated
+```
+
+## Output
+
+```
+=== init-deep Complete ===
+
+Mode: {update | create-new}
+Scope: {scope}
+Max Depth: {maxDepth}
+
+Files:
+  ✓ ./AGENTS.md (root, {N} lines)
+  ✓ ./src/hooks/AGENTS.md ({N} lines)
+  ✓ ./src/api/AGENTS.md ({N} lines)
+
+Dirs Analyzed: {N}
+AGENTS.md Created: {N}
+AGENTS.md Updated: {N}
+AGENTS.md Skipped (low score): {N}
+
+Hierarchy:
+  ./AGENTS.md
+  ├── src/hooks/AGENTS.md
+  └── src/api/AGENTS.md
+
+Next steps:
+  - Review generated files for accuracy
+  - Run /init-context for project planning workflow
+  - Run /init-user for personalization
+```
+
+## Anti-Patterns
+
+| Anti-Pattern                         | Why It Fails                                          | Instead                                            |
+| ------------------------------------ | ----------------------------------------------------- | -------------------------------------------------- |
+| Running without /init                | Root AGENTS.md will be incomplete, missing tech stack | Always run /init first                             |
+| Over-documenting trivial dirs        | Noise drowns signal, wastes context tokens            | Only document dirs scoring >8                      |
+| Repeating parent content in children | Wastes tokens, creates maintenance burden             | "See ../AGENTS.md for project-wide context"        |
+| Including generic advice             | Applies everywhere, not project-specific              | Only document deviations from standard             |
+| Verbose prose                        | Consumes context tokens without adding information    | Telegraphic style — facts only                     |
+| Skipping existing content            | Loses user-written conventions and gotchas            | Always read existing first, even with --create-new |
+| Ignoring --max-depth                 | Creates deeply nested files that won't be read        | Respect the depth limit                            |
+| Not validating commands              | Stale/broken commands mislead agents                  | Run --version or --help to verify                  |
+
+---
+
+## Skill Integration Summary
+
+| Skill                            | When Used  | Purpose                               |
+| -------------------------------- | ---------- | ------------------------------------- |
+| `index-knowledge`                | Phase 2-3  | Discovery, scoring, generation        |
+| `memory-system`                  | Phase 1, 7 | Search previous runs, record findings |
+| `verification-before-completion` | Phase 7    | Validate created files                |

package/dist/template/{.opencode → opencode}/command/lfg.md

@@ -22,10 +22,29 @@ Full compound engineering cycle. One command, all four steps.
 ## Phase 0: Preflight
 
 ```bash
-br show $BEAD_ID
+br show $ARGUMENTS
 ```
 
-Read `.beads/artifacts/$BEAD_ID/` to check what artifacts exist.
+Read `.beads/artifacts/$ARGUMENTS/` to check what artifacts exist.
+
+### Worktree Detection
+
+If the bead has a tracked worktree, operate from that directory:
+
+```bash
+WORKTREE_FILE=".beads/artifacts/$ARGUMENTS/worktree.txt"
+
+if [ -f "$WORKTREE_FILE" ]; then
+  WORKTREE_PATH="$(cat "$WORKTREE_FILE")"
+  if git worktree list --porcelain | grep -F "worktree $WORKTREE_PATH" >/dev/null; then
+    echo "Using bead worktree: $WORKTREE_PATH"
+    cd "$WORKTREE_PATH"
+  else
+    echo "Tracked worktree is stale or missing: $WORKTREE_PATH"
+    echo "Ask user whether to recreate, remove tracking file, or continue in current directory."
+  fi
+fi
+```
 
 Verify:
 
@@ -50,7 +69,7 @@ Load and execute the `/plan` command for this bead:
 ```typescript
 skill({ name: "writing-plans" });
 // Run full /plan flow including Phase 0 institutional research
-// Output: .beads/artifacts/$BEAD_ID/plan.md
+// Output: .beads/artifacts/$ARGUMENTS/plan.md
 ```
 
 Checkpoint if plan has major unknowns or architecture questions. Otherwise proceed automatically.
@@ -116,7 +135,7 @@ Awaiting your decision before continuing.
 Load and run the compound command:
 
 ```typescript
-// Run /compound $BEAD_ID
+// Run /compound $ARGUMENTS
 // Extract learnings from the full cycle
 // Store observations to memory
 // Suggest AGENTS.md updates if conventions changed

package/dist/template/{.opencode → opencode}/command/plan.md

@@ -20,6 +20,7 @@ Create a detailed implementation plan with TDD steps. Optional deep-planning bet
 skill({ name: "beads" });
 skill({ name: "memory-grounding" });
 skill({ name: "writing-plans" }); // TDD plan format
+skill({ name: "research-tool-routing" }); // Canonical Context7/Webclaw/optional Exa policy
 // For user-facing UI work: skill({ name: "ux-quality-gates" });
 ```
 
@@ -111,9 +112,9 @@ Before research, determine discovery level based on PRD:
 | Level | Scope                | When to Use                                                       | Action                                      |
 | ----- | -------------------- | ----------------------------------------------------------------- | ------------------------------------------- |
 | **0** | Skip                 | Pure internal work, existing patterns only (grep confirms)        | Skip research, proceed to decomposition     |
-| **1** | Quick (2-5 min)      | Single known library, confirming syntax/version                   | `context7 resolve-library-id + query-docs`  |
-| **2** | Standard (15-30 min) | Choosing between 2-3 options, new external integration            | Spawn `@scout` for research                 |
-| **3** | Deep (1+ hour)       | Architectural decision, novel problem, multiple external services | Full research with parallel `@scout` agents |
+| **1** | Quick (2-5 min)      | Single known library, confirming syntax/version                   | `context7 resolve-library-id + query-docs` per `research-tool-routing` |
+| **2** | Standard (15-30 min) | Choosing between 2-3 options, new external integration            | Spawn `@scout`; use Webclaw for URLs/current web content |
+| **3** | Deep (1+ hour)       | Architectural decision, novel problem, multiple external services | Full research with parallel `@scout` agents; optional Exa only for power-preset semantic/deep search |
 
 **Depth indicators:**
 
@@ -152,7 +153,7 @@ Spawn parallel agents to gather implementation context:
 | Agent     | Purpose                                                              |
 | --------- | -------------------------------------------------------------------- |
 | `explore` | Codebase patterns, affected file structure, test patterns, conflicts |
-| `scout`   | Best practices, common patterns, pitfalls                            |
+| `scout`   | Best practices, common patterns, pitfalls; apply [research-tool-routing](../skill/research-tool-routing/SKILL.md) for Context7/Webclaw/optional Exa decisions |
 
 ## Phase 4: Goal-Backward Analysis
 

package/dist/template/{.opencode → opencode}/command/pr.md

@@ -24,6 +24,28 @@ skill({ name: "verification-before-completion" });
 
 ## Phase 1: Pre-PR Verification
 
+### Worktree Detection
+
+If a bead ID is provided, check for a tracked worktree:
+
+```bash
+if [ -n "$ARGUMENTS" ]; then
+  WORKTREE_FILE=".beads/artifacts/$ARGUMENTS/worktree.txt"
+  if [ -f "$WORKTREE_FILE" ]; then
+    WORKTREE_PATH="$(cat "$WORKTREE_FILE")"
+    if git worktree list --porcelain | grep -F "worktree $WORKTREE_PATH" >/dev/null; then
+      echo "Using bead worktree: $WORKTREE_PATH"
+      cd "$WORKTREE_PATH"
+    else
+      echo "Tracked worktree is stale or missing: $WORKTREE_PATH"
+      echo "Continuing in current directory."
+    fi
+  fi
+fi
+```
+
+All subsequent PR operations must run from the correct workspace (worktree or main checkout).
+
 ```bash
 git status --porcelain
 ```
@@ -156,7 +178,8 @@ Report:
 1. PR URL
 2. Status (Ready for Review / Draft)
 3. Branch → main
-4. Gate results
+4. Workspace: [worktree path | main checkout]
+5. Gate results
 
 ## Related Commands
 

package/dist/template/{.opencode → opencode}/command/research.md

@@ -15,6 +15,7 @@ Gather information before implementation. Find answers, document findings, stop
 ```typescript
 skill({ name: "beads" });
 skill({ name: "memory-grounding" });
+skill({ name: "research-tool-routing" }); // Canonical Context7/Webclaw/optional Exa policy
 // For --thorough mode:
 skill({ name: "deep-research" });
 ```
@@ -51,9 +52,11 @@ Default depth: ~30 tool calls for moderate exploration.
 | `explore`    | Codebase patterns, LSP analysis |
 | `scout`      | External docs, best practices   |
 | `context7`   | Official API references         |
+| `webclaw`    | URL extraction, crawling, current web content, structured page extraction |
 | `opensrc`    | Package source code inspection  |
 | `codesearch` | Real-world usage examples       |
 | `grepsearch` | GitHub code search              |
+| `exa`        | Optional power-preset semantic/deep web search only |
 
 ## Phase 1: Load Context
 
@@ -73,11 +76,14 @@ Follow the [memory-grounding](../skill/memory-grounding/SKILL.md) skill protocol
 
 ### Source Priority
 
+Apply [research-tool-routing](../skill/research-tool-routing/SKILL.md) when choosing between Context7, Webclaw, and optional Exa.
+
 1. **Codebase patterns** — delegate to `explore` agent for LSP analysis
-2. **Official docs** — `context7` for API references
+2. **Official docs** — `context7` for library/API references
 3. **Source code** — `npx opensrc <package>` when docs are insufficient
 4. **GitHub examples** — `codesearch` / `grepsearch` for real-world patterns
-5. **Web search** — only if tiers 1-4 don't answer
+5. **Current web content / URLs** — `webclaw` for scrape, crawl, map, batch, extract, diff, summarize, and brand workflows
+6. **Semantic/deep web search** — optional `exa` power preset only; do not require Exa for default/recommended workflows
 
 ### Delegation
 

package/dist/template/{.opencode → opencode}/command/resume.md

@@ -32,6 +32,38 @@ git status --porcelain
 
 If not on the right branch, check out the feature branch. If uncommitted changes exist, ask user what to do.
 
+### Worktree Restoration
+
+If the bead has a tracked worktree, restore it:
+
+```bash
+WORKTREE_FILE=".beads/artifacts/$ARGUMENTS/worktree.txt"
+
+if [ -f "$WORKTREE_FILE" ]; then
+  WORKTREE_PATH="$(cat "$WORKTREE_FILE")"
+  if git worktree list --porcelain | grep -F "worktree $WORKTREE_PATH" >/dev/null; then
+    echo "Restoring bead worktree: $WORKTREE_PATH"
+    cd "$WORKTREE_PATH"
+  else
+    echo "Tracked worktree is stale or missing: $WORKTREE_PATH"
+    echo "Options:"
+    echo "  1. Recreate worktree (git worktree add <path> -b <branch>)"
+    echo "  2. Remove stale tracking and use current workspace"
+    echo "  3. Stop"
+  fi
+fi
+```
+
+Report workspace mode:
+
+```
+Workspace:
+- Mode: [worktree | branch | current]
+- Branch: [branch-name]
+- Directory: [working-directory]
+- Worktree tracking: [valid | stale | none]
+```
+
 ## Phase 3: Find Handoff
 
 Follow the [memory-grounding](../skill/memory-grounding/SKILL.md) skill protocol. Focus on: handoff file by bead ID, session history.

package/dist/template/{.opencode → opencode}/command/ship.md

@@ -57,6 +57,25 @@ skill({ name: "reflection-checkpoints" }); // Mid-point + completion checks duri
 
 Follow the [memory-grounding](../skill/memory-grounding/SKILL.md) skill protocol. Focus on: failed approaches to avoid repeating.
 
+### Worktree Detection
+
+If the bead has a tracked worktree, operate from that directory:
+
+```bash
+WORKTREE_FILE=".beads/artifacts/$ARGUMENTS/worktree.txt"
+
+if [ -f "$WORKTREE_FILE" ]; then
+  WORKTREE_PATH="$(cat "$WORKTREE_FILE")"
+  if git worktree list --porcelain | grep -F "worktree $WORKTREE_PATH" >/dev/null; then
+    echo "Using bead worktree: $WORKTREE_PATH"
+    cd "$WORKTREE_PATH"
+  else
+    echo "Tracked worktree is stale or missing: $WORKTREE_PATH"
+    echo "Ask user whether to recreate, remove tracking file, or continue in current directory."
+  fi
+fi
+```
+
 ### Bead Validation
 
 ```bash

package/dist/template/{.opencode → opencode}/command/status.md

@@ -61,6 +61,26 @@ br ready
 git status --porcelain
 git branch --show-current
 git log --oneline -5
+git worktree list
+```
+
+Show active worktrees with validation (in-progress beads only):
+
+```bash
+# Get list of in-progress bead IDs (br list outputs: ◐ bead-id [● P0] [type] - title)
+IN_PROGRESS=$(br list --status in_progress | awk '{print $2}')
+
+for BEAD_ID in $IN_PROGRESS; do
+  BEAD_DIR=".beads/artifacts/$BEAD_ID/"
+  if [ -d "$BEAD_DIR" ] && [ -f "$BEAD_DIR/worktree.txt" ]; then
+    WT_PATH=$(cat "$BEAD_DIR/worktree.txt")
+    if git worktree list --porcelain | grep -F "worktree $WT_PATH" >/dev/null; then
+      echo "  $BEAD_ID → $WT_PATH (valid)"
+    else
+      echo "  $BEAD_ID → $WT_PATH (stale)"
+    fi
+  fi
+done
 ```
 
 Optionally include Portless local URL state only if the binary is already installed:
@@ -95,6 +115,10 @@ GIT
   Changes: [from git status, or "clean"]
   Recent:  [from git log]
 
+WORKTREES
+  [list from git worktree list, or "none"]
+  Tracked: [bead worktree mappings, or "none"]
+
 SESSIONS TODAY
   [from find_sessions]
 ```

package/dist/template/{.opencode → opencode}/command/ui-review.md

@@ -2,7 +2,6 @@
 description: Review UI/UX design for quality, aesthetics, and best practices
 argument-hint: "<image-or-component-path> [--responsive] [--dark-mode]"
 agent: vision
-model: proxypal/gemini-3-pro-preview
 ---
 
 # UI Review: $ARGUMENTS

package/dist/template/{.opencode → opencode}/command/ui-slop-check.md

@@ -2,7 +2,6 @@
 description: Audit changed UI files for AI slop patterns and design-system violations
 argument-hint: "[path|auto] [--staged] [--since=<ref>] [--full-report]"
 agent: vision
-model: proxypal/gemini-3-pro-preview
 ---
 
 # UI Slop Check: $ARGUMENTS

package/dist/template/{.opencode → opencode}/command/verify.md

@@ -1,7 +1,7 @@
 ---
 description: Verify implementation completeness, correctness, and coherence
 argument-hint: "<bead-id> [--quick] [--full] [--fix] [--no-cache]"
-agent: review
+agent: verify
 ---
 
 # Verify: $ARGUMENTS

package/dist/template/{.opencode → opencode}/dcp.jsonc

@@ -62,7 +62,9 @@
 			// GPT/Gemini: assume similar 200k+ windows
 			"github-copilot/gpt-5.4": 110000,
 			"github-copilot/gpt-5.3-codex": 110000,
-			"github-copilot/gemini-3.1-pro-preview": 110000
+			"github-copilot/gemini-3.1-pro-preview": 110000,
+			// OpenCode default: big-pickle (~200k context)
+			"opencode/big-pickle": 110000
 		},
 		// Soft lower threshold: below this, turn/iteration reminders are off
 		// Use numeric values — same reason as maxContextLimit above
@@ -77,7 +79,8 @@
 			"github-copilot/claude-haiku-4.5": 50000,
 			"github-copilot/gpt-5.4": 65000,
 			"github-copilot/gpt-5.3-codex": 65000,
-			"github-copilot/gemini-3.1-pro-preview": 65000
+			"github-copilot/gemini-3.1-pro-preview": 65000,
+			"opencode/big-pickle": 65000
 		},
 		// How often context-limit nudge fires above maxContextLimit (1 = every fetch)
 		"nudgeFrequency": 5,