oh-my-customcode 0.64.1 → 0.64.3

package/dist/cli/index.js

@@ -9325,7 +9325,7 @@ var init_package = __esm(() => {
     workspaces: [
       "packages/*"
     ],
-    version: "0.64.1",
+    version: "0.64.3",
     description: "Batteries-included agent harness for Claude Code",
     type: "module",
     bin: {

package/dist/index.js

@@ -1674,7 +1674,7 @@ var package_default = {
   workspaces: [
     "packages/*"
   ],
-  version: "0.64.1",
+  version: "0.64.3",
   description: "Batteries-included agent harness for Claude Code",
   type: "module",
   bin: {

package/package.json

@@ -3,7 +3,7 @@
   "workspaces": [
     "packages/*"
   ],
-  "version": "0.64.1",
+  "version": "0.64.3",
   "description": "Batteries-included agent harness for Claude Code",
   "type": "module",
   "bin": {

package/templates/.claude/agents/arch-documenter.md

@@ -16,6 +16,7 @@ tools:
   - Glob
 maxTurns: 20
 disallowedTools: [Bash]
+permissionMode: bypassPermissions
 ---
 
 You handle software architecture documentation: system design docs, API specs, ADRs, and technical doc maintenance.

package/templates/.claude/agents/arch-speckit-agent.md

@@ -16,6 +16,7 @@ maxTurns: 20
 limitations:
   - "cannot execute code"
   - "cannot deploy infrastructure"
+permissionMode: bypassPermissions
 ---
 
 You are a Spec-Driven Development agent that transforms requirements into executable specifications.

package/templates/.claude/agents/be-django-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Django developer specialized in building production-ready Python web applications following best practices and modern patterns.

package/templates/.claude/agents/be-express-expert.md

@@ -12,6 +12,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Express.js developer for production-ready Node.js APIs following security best practices and 12-factor app principles.

package/templates/.claude/agents/be-fastapi-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert FastAPI developer specialized in building high-performance async Python APIs following best practices and modern patterns.

package/templates/.claude/agents/be-go-backend-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Go backend developer specialized in building production-ready services following Uber style guide and standard project layout.

package/templates/.claude/agents/be-nestjs-expert.md

@@ -12,6 +12,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert NestJS developer for scalable Node.js applications using TypeScript with enterprise-grade patterns.

package/templates/.claude/agents/be-springboot-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Spring Boot developer for enterprise-grade Java/Kotlin applications. Focused on Spring Boot 3.5.x with Java 21.

package/templates/.claude/agents/db-alembic-expert.md

@@ -23,6 +23,7 @@ limitations:
   - "cannot apply migrations directly to production databases"
   - "cannot resolve application-level data backfill logic without domain context"
   - "cannot detect rename intent without git diff context or explicit user instruction"
+permissionMode: bypassPermissions
 ---
 
 # db-alembic-expert

package/templates/.claude/agents/db-postgres-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert PostgreSQL DBA specialized in designing, optimizing, and maintaining pure PostgreSQL databases in production.

package/templates/.claude/agents/db-redis-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Redis developer specialized in high-performance caching, in-memory data architectures, and real-time messaging systems.

package/templates/.claude/agents/db-supabase-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert in Supabase and PostgreSQL for performant, secure database-driven applications.

package/templates/.claude/agents/de-airflow-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Apache Airflow developer for production-ready DAGs following official best practices.

package/templates/.claude/agents/de-dbt-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert dbt developer for analytics engineering, SQL modeling, and data transformation.

package/templates/.claude/agents/de-kafka-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Apache Kafka developer for event streaming architectures with high throughput and reliability.

package/templates/.claude/agents/de-pipeline-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert data pipeline architect for robust, scalable data pipelines integrating multiple tools with data quality assurance.

package/templates/.claude/agents/de-snowflake-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Snowflake developer for cloud data warehouse design, query optimization, and scalable data platforms.

package/templates/.claude/agents/de-spark-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Apache Spark developer for performant distributed data processing using PySpark and Scala.

package/templates/.claude/agents/fe-design-expert.md

@@ -14,6 +14,7 @@ disallowedTools: [Bash]
 limitations:
   - "cannot modify backend code"
   - "cannot execute shell commands"
+permissionMode: bypassPermissions
 source:
   type: external
   origin: github

package/templates/.claude/agents/fe-flutter-agent.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Flutter developer following official documentation and Dart best practices.

package/templates/.claude/agents/fe-svelte-agent.md

@@ -15,6 +15,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Svelte developer following official documentation and compiler-based reactivity patterns.

package/templates/.claude/agents/fe-vercel-agent.md

@@ -17,6 +17,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are a frontend specialist for React/Next.js projects with Vercel deployment capabilities.

package/templates/.claude/agents/fe-vuejs-agent.md

@@ -15,6 +15,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Vue.js (Vue 3) developer following official documentation and best practices.

package/templates/.claude/agents/infra-aws-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert AWS cloud architect specialized in designing and implementing scalable, secure, and cost-effective cloud infrastructure following AWS Well-Architected Framework.

package/templates/.claude/agents/infra-docker-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Docker engineer specialized in building optimized container images and managing containerized applications following official best practices.

package/templates/.claude/agents/lang-golang-expert.md

@@ -15,6 +15,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Go developer specialized in writing idiomatic, performant, and maintainable Go code following official best practices.

package/templates/.claude/agents/lang-java21-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Java 21 developer for modern Java features including Virtual Threads, Pattern Matching, Record Patterns, and Sequenced Collections.

package/templates/.claude/agents/lang-kotlin-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Kotlin developer specialized in writing idiomatic, concise, and safe Kotlin code following JetBrains official conventions.

package/templates/.claude/agents/lang-python-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Python developer specialized in writing Pythonic, clean, and maintainable code following PEP 8 and The Zen of Python.

package/templates/.claude/agents/lang-rust-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Rust developer specialized in writing safe, performant, and idiomatic Rust code following official guidelines and community best practices.

package/templates/.claude/agents/lang-typescript-expert.md

@@ -14,6 +14,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert TypeScript developer specialized in writing type-safe, maintainable, and scalable TypeScript code following industry best practices.

package/templates/.claude/agents/mgr-claude-code-bible.md

@@ -13,6 +13,7 @@ tools:
   - Write
   - Grep
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are the authoritative source of truth for Claude Code specifications. You fetch official documentation from code.claude.com and validate the project against official specs.

package/templates/.claude/agents/mgr-creator.md

@@ -15,6 +15,7 @@ tools:
   - Glob
   - Bash
 maxTurns: 25
+permissionMode: bypassPermissions
 ---
 
 You are an agent creation specialist following R006 (MUST-agent-design.md) rules.

package/templates/.claude/agents/mgr-gitnerd.md

@@ -16,6 +16,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are a Git operations specialist following GitHub flow best practices.

package/templates/.claude/agents/mgr-sauron.md

@@ -15,6 +15,7 @@ tools:
   - Glob
   - Bash
 maxTurns: 25
+permissionMode: bypassPermissions
 ---
 
 You are an automated verification specialist that executes the mandatory R017 verification process, acting as the "all-seeing eye" that ensures system integrity through comprehensive multi-round verification.

package/templates/.claude/agents/mgr-supplier.md

@@ -16,6 +16,7 @@ tools:
   - Read
   - Grep
   - Glob
+permissionMode: default
 ---
 
 You are a dependency validation specialist ensuring agents have all required skills and guides properly linked.

package/templates/.claude/agents/mgr-updater.md

@@ -19,6 +19,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an external source synchronization specialist keeping external components up-to-date.

package/templates/.claude/agents/qa-engineer.md

@@ -15,6 +15,7 @@ tools:
   - Grep
   - Glob
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are a QA execution specialist that runs tests, identifies defects, and validates software quality.

package/templates/.claude/agents/qa-planner.md

@@ -16,6 +16,7 @@ tools:
   - Edit
   - Grep
   - Glob
+permissionMode: bypassPermissions
 ---
 
 You are a QA planning specialist creating comprehensive test strategies from requirements.

package/templates/.claude/agents/qa-writer.md

@@ -16,6 +16,7 @@ tools:
   - Edit
   - Grep
   - Glob
+permissionMode: bypassPermissions
 ---
 
 You are a QA documentation specialist transforming test plans into detailed, executable test cases and reports.

package/templates/.claude/agents/sec-codeql-expert.md

@@ -14,6 +14,7 @@ tools:
   - Write
   - Grep
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are a security-focused code analyst specializing in CodeQL-based vulnerability detection and assessment.

package/templates/.claude/agents/sys-memory-keeper.md

@@ -20,6 +20,7 @@ maxTurns: 15
 limitations:
   - "cannot modify source code"
   - "cannot execute tests"
+permissionMode: bypassPermissions
 ---
 
 You are a session memory management specialist ensuring context survives across session compactions using claude-mem.

package/templates/.claude/agents/sys-naggy.md

@@ -15,6 +15,7 @@ tools:
   - Write
   - Edit
   - Grep
+permissionMode: bypassPermissions
 ---
 
 You are a task management specialist that proactively manages TODO items and reminds users of pending tasks.

package/templates/.claude/agents/tool-bun-expert.md

@@ -11,6 +11,7 @@ tools:
   - Edit
   - Grep
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You are an expert Bun runtime developer for high-performance JavaScript/TypeScript applications.

package/templates/.claude/agents/tool-npm-expert.md

@@ -15,6 +15,7 @@ tools:
   - Edit
   - Grep
   - Bash
+permissionMode: bypassPermissions
 ---
 
 You manage npm package publishing, versioning, and registry operations.

package/templates/.claude/agents/tool-optimizer.md

@@ -17,6 +17,7 @@ tools:
 maxTurns: 20
 limitations:
   - "cannot modify source code"
+permissionMode: bypassPermissions
 ---
 
 You analyze and optimize application bundles, detect performance issues, and provide actionable recommendations.

package/templates/.claude/skills/evaluator-optimizer/SKILL.md

@@ -54,6 +54,56 @@ When enabled:
 
 Use when: tasks requiring 3+ iterations consistently, or when generator-evaluator score disagreements exceed 0.3.
 
+### Evaluator Calibration
+
+Anthropic's harness design research identifies evaluator leniency as a key failure mode: LLMs default to generous scoring, especially when evaluating output from the same model family. Counter-measures:
+
+**Skepticism Prompting**: Include explicit instructions in the evaluator prompt:
+- "Default to skepticism. A 'pass' should require clear evidence, not absence of issues."
+- "Score as if you are reviewing code that will run in production with real users."
+- "When uncertain between pass and fail, choose fail and explain what evidence would change your mind."
+
+**Anti-Self-Praise Bias**: When generator and evaluator share the same model family (e.g., both Claude), add:
+- "You are reviewing another agent's work, not your own. Do not give credit for intent — only for execution."
+- "Identify at least one concrete improvement, even for high-quality output."
+
+**Calibration via Rubric Examples**: Each rubric criterion SHOULD include a `fail_example` alongside the description:
+
+```yaml
+rubric:
+  - criterion: error_handling
+    weight: 0.25
+    description: "All error paths handled with meaningful messages"
+    fail_example: "Generic try/catch with console.log(error) — no recovery, no user-facing message"
+```
+
+Adding `fail_example` anchors the evaluator's scale, reducing score inflation by ~20% (based on Anthropic's internal testing).
+
+### Conditional Evaluator (Cost Optimization)
+
+Not every task justifies evaluator overhead. Skip the evaluator loop for tasks within the model's reliable capability range. From Anthropic's research: "Worth cost when tasks sit beyond baseline model capability; unnecessary overhead for problems within model's reliable range."
+
+```yaml
+evaluator-optimizer:
+  conditional:
+    enabled: true
+    skip_when:
+      - task_complexity: low        # Simple, well-defined tasks
+      - generator_confidence: high  # Generator self-reports high confidence
+      - historical_pass_rate: 0.9   # Same task type historically passes first try
+```
+
+When `conditional.enabled: true` and ANY `skip_when` condition is met, the evaluator is skipped and the generator's first output is returned directly. This reduces token cost by ~40% for straightforward tasks.
+
+**Decision matrix**:
+
+| Task Type | Complexity | Evaluator? |
+|-----------|-----------|------------|
+| Simple file rename, config change | Low | Skip |
+| Standard CRUD implementation | Medium | Run |
+| Complex architecture, security-critical | High | Run with pre-negotiation |
+| Previously failed task retry | Any | Always run |
+
 ### Parameter Details
 
 | Parameter | Required | Default | Description |
@@ -224,6 +274,7 @@ evaluator-optimizer:
     - criterion: correctness
       weight: 0.35
       description: Code compiles, logic is correct, edge cases handled
+      fail_example: "Missing null check on user input causes runtime crash"
     - criterion: style
       weight: 0.2
       description: Follows project conventions, clean and readable
@@ -328,6 +379,7 @@ When ecomode is active (R013), compress output:
 - The evaluator prompt MUST include the full rubric to ensure consistent scoring
 - Iteration state (best score, best output) is tracked by the orchestrator
 - The hard cap of 5 iterations prevents runaway refinement loops
+- For multi-sprint runs (5+ iterations), consider context reset: spawn a fresh evaluator agent rather than continuing with degraded context. The workflow-runner supports this via `context: fork` on individual steps. Anthropic's research confirms "context resets provide clean slates superior to compaction" for long-running evaluation.
 
 ## Domain Examples
 

package/templates/manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "0.64.1",
+  "version": "0.64.3",
   "lastUpdated": "2026-03-24T00:00:00.000Z",
   "components": [
     {