bmalph 2.7.5 → 2.7.6

package/ralph/templates/PROMPT.md

@@ -34,6 +34,11 @@ You are Ralph, an autonomous AI development agent working on a [YOUR PROJECT NAM
 - Keep .ralph/@AGENT.md updated with build/run instructions
 - Document the WHY behind tests and implementations
 - No placeholder implementations - build it properly
+- do not ask the user questions during loop execution
+- do not use AskUserQuestion, EnterPlanMode, or ExitPlanMode during loop execution
+- make the safest reasonable assumption and continue
+- prefer small, reversible changes when requirements are ambiguous
+- surface blockers in the Ralph status block instead of starting a conversation
 
 ## 🎯 Status Reporting (CRITICAL - Ralph needs this!)
 

package/ralph/templates/ralphrc.template

@@ -40,10 +40,20 @@ CLAUDE_OUTPUT_FORMAT="json"
 # TOOL PERMISSIONS
 # =============================================================================
 
-# Comma-separated list of allowed tools for Claude
-# Common tools: Write, Read, Edit, Grep, Glob
-# Bash patterns: Bash(git *), Bash(npm *), Bash(pytest)
-ALLOWED_TOOLS="Write,Read,Edit,Bash(git *),Bash(npm *),Bash(pytest)"
+# Comma-separated list of allowed tools for Claude Code only.
+# Ignored by the codex, cursor, and copilot drivers.
+# Opt in to interactive pauses by adding AskUserQuestion manually.
+ALLOWED_TOOLS="Write,Read,Edit,MultiEdit,Glob,Grep,Task,TodoWrite,WebFetch,WebSearch,NotebookEdit,Bash"
+
+# Permission mode for Claude Code CLI (default: auto)
+# Options: auto, acceptEdits, bypassPermissions, default, dontAsk, plan
+CLAUDE_PERMISSION_MODE="auto"
+
+# How Ralph responds when a driver reports permission denials:
+# - continue: log the denial and keep looping (default for unattended mode)
+# - halt: stop immediately and show recovery guidance
+# - threshold: continue until the permission-denial circuit breaker trips
+PERMISSION_DENIAL_MODE="continue"
 
 # =============================================================================
 # SESSION MANAGEMENT