agent-bober 0.4.2 → 0.4.3

package/README.md

@@ -101,14 +101,40 @@ Specialized workflows:
 ### CLI
 
 ```bash
-npx agent-bober init [preset]       # Initialize project (nextjs, react-vite, solidity, anchor, api-node, python-api, brownfield)
-npx agent-bober plan                # Run the planner
+npx agent-bober init [preset]       # Initialize project
+npx agent-bober plan "feature"      # Run the planner
 npx agent-bober sprint              # Execute next sprint
 npx agent-bober eval                # Evaluate current sprint
-npx agent-bober run                 # Full autonomous loop
-npx agent-bober status              # Show plan progress
+npx agent-bober run "feature"       # Full autonomous loop
 ```
 
+### Fully Autonomous Mode (no human in the loop)
+
+**Option A: Claude Code (recommended)**
+
+Launch Claude Code with auto-accept permissions, then run the pipeline:
+
+```bash
+cd your-project
+agent-bober init nextjs
+claude --dangerously-skip-permissions
+# Inside Claude Code:
+/bober-run Build a complete dashboard with auth, CRUD, and charts
+```
+
+Claude will plan, build, evaluate, rework, and iterate without asking you anything. Come back to a finished project.
+
+**Option B: CLI with API key**
+
+```bash
+export ANTHROPIC_API_KEY=sk-ant-...
+cd your-project
+agent-bober init nextjs
+agent-bober run "Build a complete dashboard with auth, CRUD, and charts"
+```
+
+The CLI uses the Anthropic SDK directly — no approval prompts at all.
+
 ---
 
 ## Configuration

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-bober",
-  "version": "0.4.2",
+  "version": "0.4.3",
   "description": "Generator-Evaluator multi-agent harness for building applications autonomously with Claude. Implements planner, sprint, and evaluator patterns.",
   "type": "module",
   "main": "dist/index.js",

package/skills/bober.run/SKILL.md

@@ -8,6 +8,18 @@ argument-hint: <task-description>
 
 You are running the **bober.run** skill. This is the top-level orchestrator that runs the entire Generator-Evaluator pipeline from start to finish: planning, sprint execution, evaluation, and iteration. The user provides a task description and you deliver a working implementation.
 
+## Autonomous Mode
+
+This command is designed to run **fully autonomously** — do NOT stop to ask the user for confirmation between phases unless something is genuinely ambiguous or blocked. Specifically:
+
+- **Do NOT ask** "should I continue to the next sprint?" — just continue.
+- **Do NOT ask** "should I start building?" after planning — just start.
+- **Do NOT ask** "should I rework?" after a failed evaluation — just rework (up to maxIterations).
+- **Do NOT ask** for approval on file writes, commits, or evaluation runs — just do them.
+- **DO stop** only if: you hit maxIterations on a sprint and cannot progress, or the task description is genuinely unclear and you cannot infer intent.
+
+The user launched this command to walk away and come back to a finished product. Respect that intent.
+
 ## Overview
 
 The pipeline follows this flow:
@@ -53,14 +65,9 @@ If `bober.config.json` exists, read the configuration.
 
 Read `.bober/specs/` and `.bober/progress.md`. If there is an existing plan with incomplete sprints:
 
-Ask the user:
-```
-I found an existing plan: "<plan title>" with <N> sprints (<M> completed, <K> remaining).
-
-A) Continue with the existing plan (resume from sprint <next>)
-B) Create a new plan for your task (the existing plan stays but won't be executed)
-C) Archive the existing plan and start fresh
-```
+- If the user provided a new task description that clearly differs from the existing plan → create a new plan (option B)
+- If the user provided no task or a task that matches the existing plan → resume from the next incomplete sprint (option A)
+- Log your decision but do NOT ask the user — autonomous mode means you decide and move forward
 
 ### 1c. Run the Planning Phase