cralph 1.0.0-beta.6 → 1.0.0-beta.7

package/README.md

@@ -62,8 +62,9 @@ cralph --yes
 2. Looks for `.ralph/` in current directory only (not subdirectories)
 3. Loads config from `.ralph/paths.json` or creates starter structure
 4. Runs `claude -p --dangerously-skip-permissions` in a loop
-5. Claude updates `.ralph/TODO.md` after each iteration
-6. Stops when Claude outputs `<promise>COMPLETE</promise>`
+5. Claude completes **ONE task per iteration**, marks it done, then stops
+6. Auto-commits progress after each iteration (fails gracefully if no git)
+7. Stops when Claude outputs `<promise>COMPLETE</promise>`
 
 ## Config
 
@@ -90,7 +91,7 @@ Save as `.ralph/paths.json`. Refs are optional reference material (read-only).
 
 ### TODO Format
 
-Claude maintains this structure:
+Claude maintains this structure (one task per iteration):
 
 ```markdown
 # Tasks
@@ -102,10 +103,13 @@ Claude maintains this structure:
 
 # Notes
 
-## Task Title - Done
+## Task 1 - Done
 - What was implemented
 - Files changed
 - Learnings: patterns discovered, gotchas encountered
+
+## Task 2 - Done
+- ...
 ```
 
 ## First Run (No .ralph/ in cwd)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cralph",
-  "version": "1.0.0-beta.6",
+  "version": "1.0.0-beta.7",
   "description": "Claude in a loop. Point at refs, give it a rule, let it cook.",
   "author": "mguleryuz",
   "license": "MIT",

package/src/prompt.ts

@@ -8,33 +8,22 @@ const BASE_PROMPT = `You are an autonomous coding agent running in a loop.
 
 FIRST: Read and internalize the rules provided below.
 
-## Your Task Each Iteration
+## Your Task This Iteration
 
 1. Read the TODO file
 2. Pick the FIRST uncompleted task (marked with [ ])
 3. Implement that SINGLE task
 4. Run quality checks (typecheck, lint, test - whatever the project requires)
 5. If checks pass, mark the task [x] complete
-6. Append your progress to the Notes section (see format below)
-7. If ALL tasks are complete, output the completion signal
+6. Append your progress to the Notes section
+7. **STOP** - End your response. Another iteration will handle the next task.
 
 ## Critical Rules
 
-- **ONE task per iteration** - Do not try to complete multiple tasks
+- **ONE task per iteration** - Complete exactly ONE task, then STOP. Do NOT continue to the next task.
 - **Quality first** - Do NOT mark a task complete if tests/typecheck fail
 - **Keep changes focused** - Minimal, targeted changes only
 - **Follow existing patterns** - Match the codebase style
-- **Commit after completing** - Commit your changes with a meaningful message
-
-## Commit Format
-
-After completing a task and quality checks pass, commit ALL changes with:
-
-\`\`\`
-feat: [Task Title]
-\`\`\`
-
-Use \`feat:\` for new features, \`fix:\` for bug fixes, \`refactor:\` for refactoring, \`docs:\` for documentation.
 
 ## Progress Format
 
@@ -55,13 +44,15 @@ If refs paths are provided, they are READ-ONLY reference material. Never modify
 
 ## Stop Condition
 
-After completing a task, check if ALL tasks are marked [x] complete.
+After completing ONE task, check the TODO file:
+
+- If there are still tasks marked [ ] (pending): **END your response normally.** Another iteration will pick up the next task.
 
-If ALL tasks are done, output exactly:
+- If ALL tasks are marked [x] (complete): Output exactly:
 
 <promise>COMPLETE</promise>
 
-If there are still pending tasks, end your response normally (the loop will continue).`;
+**IMPORTANT:** Do NOT continue to the next task. Complete ONE task, then STOP.`;
 
 /**
  * Build the complete prompt with config and rules injected