cortex-agents 4.0.6 → 4.0.8

package/.opencode/agents/coder.md

@@ -1,5 +1,5 @@
 ---
-description: Multi-layer code implementation across frontend, backend, and database
+description: Code implementation sub-agent — handles all task types
 mode: subagent
 temperature: 0.3
 tools:
@@ -13,7 +13,7 @@ permission:
   bash: ask
 ---
 
-You are a fullstack developer. You implement complete features spanning frontend, backend, and database layers with consistent contracts across the stack.
+You are a skilled developer. You implement tasks ranging from single-file changes to full-stack features spanning frontend, backend, and database layers with consistent contracts across the stack.
 
 ## Auto-Load Skills (based on affected layers)
 
@@ -37,12 +37,14 @@ You are launched as a sub-agent by a primary agent in one of two contexts:
 
 ### Context A — Implementation (from implement agent)
 
-You receive requirements and implement end-to-end features across multiple layers. You will get:
-- The plan or requirements describing the feature
-- Current codebase structure for relevant layers
-- Any API contracts or interfaces that need to be consistent across layers
+You receive requirements and implement the task. Tasks range from single-file changes to end-to-end features across multiple layers. You will get:
+- The task title, description, and acceptance criteria
+- Relevant files and codebase structure
+- Build/test commands for verification
+- Files modified by previous tasks (for cross-task consistency)
+- On retries: error output from failed build/test and summary of previous attempt
 
-**Your job:** Implement the feature across all affected layers, maintaining consistency. Write the code, ensure interfaces match, and return a structured summary.
+**Your job:** Implement the task, write the code, ensure interfaces match (for multi-layer tasks), and return a structured summary.
 
 ### Context B — Feasibility Analysis (from architect agent)
 
@@ -57,6 +59,24 @@ You receive requirements and analyze implementation feasibility. You will get:
 
 ### For Context A (Implementation)
 
+Use the format that matches the task scope:
+
+**Single-layer / focused tasks:**
+```
+### Implementation Summary
+- **Files created**: [count]
+- **Files modified**: [count]
+
+### Changes
+- `path/to/file.ts` — [what was done]
+- `path/to/file.test.ts` — [what was done]
+
+### Notes
+- [Any assumptions made]
+- [Things the orchestrating agent should verify]
+```
+
+**Multi-layer / full-stack tasks:**
 ```
 ### Implementation Summary
 - **Layers modified**: [frontend, backend, database, infrastructure]

package/.opencode/agents/implement.md

@@ -66,7 +66,7 @@ permission:
     "npm run lint --*": allow
 ---
 
-You are an expert software developer. Your role is to write clean, maintainable, and well-tested code.
+You are an expert software development orchestrator. Your role is to analyze plans, delegate implementation tasks to the `@coder` sub-agent, verify results, and manage the development workflow. You do NOT write code directly — all code changes are performed by `@coder`.
 
 ## Pre-Implementation Workflow (MANDATORY)
 
@@ -129,9 +129,9 @@ Implement plan tasks iteratively using the REPL loop. Each task goes through a *
 
 **Session recovery:** Run `repl_resume` first to check for an interrupted loop from a previous session. If found, it will show progress and the interrupted task — skip to 6b to continue.
 
-**If no plan was loaded in Step 3**, fall back to implementing changes directly (skip to 6c without the loop tools) and proceed to Step 7 when done.
+**If no plan was loaded in Step 3**, delegate the user's request to `@coder` via the Task tool (skip to 6c without the loop tools) and proceed to Step 7 when done.
 
-**Multi-layer feature detection:** If the task involves changes across 3+ layers (e.g., database + API + frontend, or CLI + library + tests), launch the **@coder sub-agent** via the Task tool to implement the end-to-end feature.
+**ALL implementation tasks are delegated to `@coder`.** The implement agent does NOT write code directly. For every task, prepare context and launch `@coder` via the Task tool (see Step 6c).
 
 #### 6a: Initialize the Loop
 Run `repl_init` with the plan filename from Step 3.
@@ -140,8 +140,12 @@ Review the auto-detected build/test commands. If they look wrong, re-run with ma
 #### 6b: Check Loop Status
 Run `repl_status` to see the next pending task, current progress, build/test commands, and acceptance criteria (ACs) for the current task. Implement to satisfy all listed ACs.
 
-#### 6c: Implement the Current Task
-Read the task description and implement it. Write the code changes needed for that specific task.
+#### 6c: Delegate to @coder Sub-Agent
+Prepare context from `repl_status` output and launch `@coder` via the Task tool:
+1. **Gather context** — Task title, description, acceptance criteria, relevant files, and build/test commands from the `repl_status` output
+2. **Include cross-task context** — List files created or modified by previous tasks so `@coder` can maintain consistency
+3. **Launch `@coder`** — Pass all gathered context via the Task tool
+4. **Review the summary** — When `@coder` returns, review its implementation summary before proceeding to 6d (verification)
 
 #### 6d: Verify — Build + Test
 Run the build command (from repl_status output) via bash.
@@ -157,7 +161,7 @@ Run `repl_report` with the result:
 #### 6f: Loop Decision
 Based on the repl_report response:
 - **"Next: Task #N"** → Go to 6b (pick up next task)
-- **"Fix the issue, N retries remaining"** → Fix the code, go to 6d (re-verify)
+- **"Fix the issue, N retries remaining"** → Re-launch `@coder` with: the original task description, error output from the failed build/test, and a summary of the previous `@coder` attempt. Then go to 6d (re-verify)
 - **"ASK THE USER"** → Use the question tool:
   "Task #N has failed after 3 attempts. How would you like to proceed?"
   Options:
@@ -410,7 +414,7 @@ The following sub-agents are available via the Task tool. **Launch multiple sub-
 | `@audit` | Standard + High scope changes | Code quality, tech debt, pattern review | Step 7 — scope-based |
 | `@docs-writer` | Standard + High scope changes | Auto-generates decision/feature/flow docs | Step 7 — scope-based |
 | `@perf` | High scope or hot-path/DB/render changes | Complexity analysis, N+1 detection, bundle impact | Step 7 — conditional |
-| `@coder` | Multi-layer features (3+ layers) | End-to-end implementation across frontend/backend/database | Step 6 — conditional |
+| `@coder` | ALL implementation tasks | Code implementation for every task — single-file to full-stack | Step 6c — always |
 | `@devops` | High scope or CI/CD/Docker/infra files changed | Config validation, best practices checklist | Step 7 — conditional |
 | `@refactor` | Plan type is `refactor` | Behavior-preserving restructuring with test verification | Step 6 — conditional |
 | `@debug` | Issues found during implementation | Root cause analysis, troubleshooting | Step 6 — conditional |

package/dist/tools/worktree.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"worktree.d.ts","sourceRoot":"","sources":["../../src/tools/worktree.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAQvD,KAAK,MAAM,GAAG,WAAW,CAAC,QAAQ,CAAC,CAAC;AAEpC;;;GAGG;AACH,wBAAgB,YAAY,CAAC,MAAM,EAAE,MAAM;;;;;;;;;;;;;;;;;;;;EA0H1C;AAED,eAAO,MAAM,IAAI;;;;CAiCf,CAAC;AAEH;;;GAGG;AACH,wBAAgB,YAAY,CAAC,MAAM,EAAE,MAAM;;;;;;;;;;EA0F1C;AAED,eAAO,MAAM,IAAI;;;;;;;;CAgDf,CAAC"}
+{"version":3,"file":"worktree.d.ts","sourceRoot":"","sources":["../../src/tools/worktree.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AASvD,KAAK,MAAM,GAAG,WAAW,CAAC,QAAQ,CAAC,CAAC;AAEpC;;;GAGG;AACH,wBAAgB,YAAY,CAAC,MAAM,EAAE,MAAM;;;;;;;;;;;;;;;;;;;;EA0H1C;AAED,eAAO,MAAM,IAAI;;;;CAiCf,CAAC;AAEH;;;GAGG;AACH,wBAAgB,YAAY,CAAC,MAAM,EAAE,MAAM;;;;;;;;;;EAyG1C;AAED,eAAO,MAAM,IAAI;;;;;;;;CAgDf,CAAC"}

package/dist/tools/worktree.js

@@ -2,6 +2,7 @@ import { tool } from "@opencode-ai/plugin";
 import * as fs from "fs";
 import * as path from "path";
 import { git } from "../utils/shell.js";
+import { detectWorktreeInfo } from "../utils/worktree-detect.js";
 const WORKTREE_ROOT = ".worktrees";
 /**
  * Factory function that creates the worktree_create tool with access
@@ -175,7 +176,20 @@ export function createRemove(client) {
         },
         async execute(args, context) {
             const { name, deleteBranch = false } = args;
-            const worktreePath = path.join(context.worktree, WORKTREE_ROOT, name);
+            // Resolve the main repo root — if we're inside a worktree, context.worktree
+            // points to the worktree itself, not the main repo. We need the main repo
+            // to construct the correct .worktrees/ path and to run git commands from outside.
+            let mainRepoRoot = context.worktree;
+            try {
+                const info = await detectWorktreeInfo(context.worktree);
+                if (info.isWorktree && info.mainWorktreePath) {
+                    mainRepoRoot = info.mainWorktreePath;
+                }
+            }
+            catch {
+                // Fall back to context.worktree
+            }
+            const worktreePath = path.join(mainRepoRoot, WORKTREE_ROOT, name);
             const absoluteWorktreePath = path.resolve(worktreePath);
             // Check if worktree exists
             if (!fs.existsSync(absoluteWorktreePath)) {
@@ -192,14 +206,15 @@ Use worktree_list to see existing worktrees.`;
             catch {
                 // Ignore error, branch detection is optional
             }
-            // Remove the worktree
+            // Remove the worktree — must run from the main repo, not from inside
+            // the worktree being removed (git rejects that).
             try {
-                await git(context.worktree, "worktree", "remove", absoluteWorktreePath);
+                await git(mainRepoRoot, "worktree", "remove", absoluteWorktreePath);
             }
             catch {
                 // Try force remove if there are changes
                 try {
-                    await git(context.worktree, "worktree", "remove", "--force", absoluteWorktreePath);
+                    await git(mainRepoRoot, "worktree", "remove", "--force", absoluteWorktreePath);
                 }
                 catch (error2) {
                     try {
@@ -224,7 +239,7 @@ The worktree may have uncommitted changes. Commit or stash them first.`;
             // Delete branch if requested
             if (deleteBranch && branchName) {
                 try {
-                    await git(context.worktree, "branch", "-d", branchName);
+                    await git(mainRepoRoot, "branch", "-d", branchName);
                     output += `\n\u2713 Deleted branch ${branchName}`;
                 }
                 catch (error) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cortex-agents",
-  "version": "4.0.6",
+  "version": "4.0.8",
   "description": "Supercharge OpenCode with structured workflows, intelligent agents, and automated development practices",
   "type": "module",
   "main": "dist/index.js",