oh-my-customcodex 0.1.0

package/templates/.claude/rules/SHOULD-memory-integration.md

@@ -0,0 +1,371 @@
+# [SHOULD] Memory Integration Rules
+
+> **Priority**: SHOULD | **ID**: R011
+
+## Architecture
+
+**Primary**: Native auto memory (`memory` field in agent frontmatter). No external dependencies.
+**Supplementary**: claude-mem MCP (optional, for cross-session search and temporal queries).
+
+Rule: If native auto memory can handle it, do NOT use claude-mem.
+
+## Native Auto Memory
+
+Agent frontmatter `memory: project|user|local` enables persistent memory:
+- System creates memory directory, loads first 200 lines of MEMORY.md into prompt
+- Read/Write/Edit tools auto-enabled for memory directory
+- Custom directory: set `autoMemoryDirectory` in settings to override default paths (v2.1.74+)
+
+| Scope | Location | Git Tracked |
+|-------|----------|-------------|
+| `user` | `~/.codex/agent-memory/<name>/` | No |
+| `project` | `.codex/agent-memory/<name>/` | Yes |
+| `local` | `.codex/agent-memory-local/<name>/` | No |
+
+## When to Use claude-mem
+
+| Scenario | Native | claude-mem |
+|----------|--------|------------|
+| Agent learns project patterns | Yes | |
+| Search across sessions | | Yes |
+| Temporal queries | | Yes |
+| Cross-agent sharing | | Yes |
+
+## Best Practices
+
+- Consult memory before starting work
+- Update after discovering patterns
+- Keep MEMORY.md under 200 lines
+- Do not store sensitive data or duplicate AGENTS.md content
+- Memory write failures should not block main task
+
+<!-- DETAIL: Confidence-Tracked Memory (sys-memory-keeper reference)
+
+Memory entries in MEMORY.md should include confidence annotations to distinguish verified facts from hypotheses.
+
+### Confidence Levels
+
+| Level | Tag | Meaning | Example |
+|-------|-----|---------|---------|
+| High | `[confidence: high]` | Verified across multiple sessions or confirmed by user | Architecture decisions, confirmed patterns |
+| Medium | `[confidence: medium]` | Observed pattern, not yet fully verified | Code conventions seen in 2-3 files |
+| Low | `[confidence: low]` | Single observation or hypothesis | First-time discovery, untested assumption |
+
+### Format in MEMORY.md
+
+    ### Key Patterns [confidence: high]
+    - `.codex/` files are gitignored — always use `git add -f`
+    - pre-commit hooks auto-detect README/manifest count mismatches
+
+    ### Hypotheses [confidence: medium]
+    - Template sync might need CI enforcement (seen in 2 PRs)
+
+    ### Unverified [confidence: low]
+    - Possible race condition in parallel hook execution (observed once)
+
+### Confidence Lifecycle
+
+    [low] — observed again — [medium] — confirmed by user/testing — [high]
+    [any] — contradicted by evidence — demoted or removed
+
+### Temporal Decay
+
+Memory entries include an optional verification timestamp for decay tracking.
+
+Format: `[confidence: high, verified: 2026-03-15]`
+
+| Age (unverified) | Action |
+|-------------------|--------|
+| 0-30 days | No change — entry is fresh |
+| 30-60 days | Demote one level (high->medium, medium->low) |
+| 60-90 days | Demote again if not re-verified |
+| 90+ days | Removal candidate — flag for review |
+
+Decay Schedule:
+    Day 0:   [confidence: high, verified: 2026-03-15]
+    Day 30:  [confidence: high, verified: 2026-03-15]  <- still within window
+    Day 31:  [confidence: medium, verified: 2026-03-15] <- auto-demoted
+    Day 61:  [confidence: low, verified: 2026-03-15]    <- demoted again
+    Day 91:  [REVIEW NEEDED, verified: 2026-03-15]      <- flagged
+
+Re-verification: Any session that confirms a memory entry resets the verified date:
+    Before: [confidence: medium, verified: 2026-01-15]
+    Action: Pattern confirmed in session
+    After:  [confidence: high, verified: 2026-03-15]
+
+Enforcement: sys-memory-keeper checks decay at session start and end:
+1. Session start: scan MEMORY.md for entries past decay threshold
+2. Flag stale entries with `[STALE]` prefix
+3. Session end: remove or demote unconfirmed stale entries
+
+Exceptions: Entries marked `[permanent]` are exempt from decay:
+    ### Architecture Decisions [confidence: high, permanent]
+
+-->
+
+<!-- DETAIL: Behavioral Memory (sys-memory-keeper reference)
+
+MEMORY.md supports an optional `## Behaviors` section for tracking user interaction preferences and workflow patterns.
+
+### Behaviors Section Format
+
+    ## Behaviors [confidence: medium]
+    - User prefers concise responses — 3 sentences max
+    - Commit messages always include issue number
+    - Security-first review perspective
+
+    ## Behavior Lifecycle
+    - New observation — [confidence: low]
+    - Seen in 2+ sessions — [confidence: medium]
+    - User-confirmed — [confidence: high]
+    - Contradicted — demote or remove
+
+### What Counts as a Behavior
+
+| Category | Examples |
+|----------|---------|
+| Communication | Verbosity preference, language, format |
+| Workflow | Tool preferences, review habits, branching patterns |
+| Domain priority | Security-first, performance-first, simplicity-first |
+
+### What Does NOT Count as a Behavior
+
+- Facts about the codebase (use existing sections)
+- One-time instructions (ephemeral, not persistent)
+- Tool configuration (belongs in AGENTS.md or settings)
+
+### Extraction Guidelines
+
+sys-memory-keeper extracts behavioral patterns at session end:
+1. Analyze conversation for repeated user preferences
+2. New behaviors start at `[confidence: low]`
+3. Promote on repeated observation across sessions
+4. Demote or remove when contradicted
+
+### Budget Management
+
+Behaviors share the 200-line MEMORY.md budget with facts. When approaching the limit:
+1. Prune `[confidence: low]` behaviors first
+2. Then prune `[confidence: medium]` behaviors
+3. `[confidence: high]` behaviors are never auto-pruned
+
+### Precedence
+
+Behavioral memory observations override soul defaults (R006 Soul Identity) when they conflict. Behaviors are user-specific and session-derived; souls are template defaults.
+
+### Rules
+
+| Rule | Detail |
+|------|--------|
+| New discoveries | Start at `[confidence: low]` unless user explicitly confirms |
+| Cross-session verification | Promote to `[confidence: medium]` when seen in 2+ sessions |
+| User confirmation | Promote to `[confidence: high]` when user confirms or tests pass |
+| Contradiction | Demote or remove when contradicted by new evidence |
+| Default | Entries without tags are treated as `[confidence: high]` (backward compatibility) |
+
+### Integration with Session-End
+
+When sys-memory-keeper updates MEMORY.md at session end:
+1. New findings from this session — `[confidence: low]`
+2. Findings that match existing entries — promote confidence
+3. Findings that contradict existing entries — flag for review
+
+-->
+
+<!-- DETAIL: Agent Metrics (sys-memory-keeper reference)
+
+MEMORY.md supports an optional `## Metrics` section for tracking per-agent-type performance data.
+
+### Metrics Section Format
+
+    ## Metrics [auto-updated by sys-memory-keeper]
+
+    | Agent Type | Tasks | Success Rate | Avg Model | Last Used |
+    |------------|-------|-------------|-----------|-----------|
+    | lang-golang-expert | 12 | 92% | sonnet | 2026-03-15 |
+    | mgr-gitnerd | 8 | 100% | sonnet | 2026-03-15 |
+
+### Metrics Collection
+
+sys-memory-keeper aggregates metrics at session end:
+
+1. Read `/tmp/.codex-task-outcomes-${PPID}` (JSONL from task-outcome-recorder hook)
+2. Parse each entry: `{agent_type, outcome, model, timestamp}`
+3. Aggregate by agent_type: total tasks, success count, model distribution
+4. Merge with existing Metrics table in MEMORY.md
+5. Budget: max 20 rows (prune lowest-usage agents when exceeded)
+
+### Metrics Fields
+
+| Field | Source | Calculation |
+|-------|--------|-------------|
+| Tasks | task-outcome-recorder JSONL | Count of entries per agent_type |
+| Success Rate | outcome field | `success_count / total_count * 100` |
+| Avg Model | model field | Most frequently used model |
+| Last Used | timestamp field | Most recent invocation |
+
+### Budget Management
+
+The Metrics section shares the 200-line MEMORY.md budget:
+1. Max 20 agent rows in Metrics table
+2. When adding new agent, prune agent with lowest task count
+3. Merge identical agent types across sessions (cumulative)
+
+-->
+
+<!-- DETAIL: User Model (sys-memory-keeper reference)
+
+MEMORY.md supports an optional `## User Model` section (DISTINCT from `## Behaviors`) for tracking structured user interaction patterns.
+
+### User Model vs Behaviors
+
+| Aspect | Behaviors | User Model |
+|--------|-----------|------------|
+| Focus | Communication/workflow preferences | Correction patterns, expertise, skill usage |
+| Source | Conversation style observations | R016 violations, tool invocations, override decisions |
+| Update | Session-end extraction | Session-end aggregation from task outcomes |
+
+### User Model Section Format
+
+    ## User Model [auto-updated by sys-memory-keeper]
+
+    ### Correction Patterns [confidence: medium]
+    - R010 direct-write attempts: 3 times → now delegates consistently
+    - Prefers explicit agent selection over auto-routing
+
+    ### Skill Preferences [confidence: high]
+    | Skill | Invocations | Last Used |
+    |-------|------------|-----------|
+    | /research | 12 | 2026-04-03 |
+    | /pipeline auto-dev | 8 | 2026-04-03 |
+
+    ### Expertise Profile [confidence: medium]
+    - Primary domains: TypeScript, Python, Go
+    - Focus: AI agent orchestration, CLI tooling
+
+    ### Override Decisions [confidence: low]
+    - Overrode /scout SKIP verdict → INTEGRATE for RTK (#756)
+
+### Categories
+
+| Category | Source | Description |
+|----------|--------|-------------|
+| Correction Patterns | R016 violation history | Rules the user corrected most |
+| Skill Preferences | Skill tool invocation count | Most-invoked skills ranked |
+| Expertise Profile | File patterns + routing history | User's domain expertise areas |
+| Override Decisions | Explicit user overrides | When user disagreed with agent recommendation |
+
+### Budget Management
+
+User Model shares the 200-line MEMORY.md budget:
+- Max 30 lines for User Model section
+- Prune low-confidence entries first
+- Skill Preferences table: max 10 rows (top by invocation count)
+- Override Decisions: max 5 entries (most recent)
+
+### Extraction Guidelines
+
+sys-memory-keeper extracts user model data at session end:
+1. Parse task outcomes for Skill invocations → update Skill Preferences
+2. Scan conversation for R016 violations → update Correction Patterns
+3. Analyze file patterns and routing decisions → update Expertise Profile
+4. Detect explicit user overrides (verdict changes, agent redirects) → update Override Decisions
+
+### Precedence
+
+User Model data feeds into intent-detection (R015) and routing skill confidence scoring. Higher expertise in a domain → higher confidence for auto-routing to that domain's agent.
+
+-->
+
+## Mid-Session Immediate Save
+
+Save memory IMMEDIATELY upon surprising discovery — do not defer to session end.
+
+| Trigger | Action | Rationale |
+|---------|--------|-----------|
+| Repeated pattern observed (2nd time) | Save `feedback_*.md` now | Pattern will recur within session |
+| Unexpected tool behavior / workaround | Save `feedback_*.md` now | Session state defense |
+| Subagent false-positive detected | Save `feedback_*.md` now | Prevent repeat in same session |
+| User correction / feedback | Save `feedback_*.md` now | Honor correction immediately |
+
+### Why Immediate?
+
+Session-end saves lose context: by the time the session ends, multiple discoveries have compounded and nuance is lost. Immediate saves preserve the exact trigger context that makes the memory actionable.
+
+**Anti-pattern**: "I'll batch all learnings at session end" — by then you'll have forgotten WHY each one mattered, and further violations may have occurred using the un-saved pattern.
+
+### Cross-reference
+
+Related records from session v0.87.2~v0.88.0 (issue #869):
+- `feedback_subagent_pre_existing_claims.md`
+- `feedback_github_workflows_inventory.md`
+- `feedback_bun_mock_module.md`
+
+## Session-End Auto-Save
+
+### Trigger
+
+Session-end detected when user says: "끝", "종료", "마무리", "done", "wrap up", "end session", or explicitly requests session save.
+
+### Flow
+
+```
+User signals session end
+  → Orchestrator delegates to sys-memory-keeper
+    → sys-memory-keeper performs:
+       1. Collect session summary (tasks, decisions, open items)
+       2. Update native auto-memory (MEMORY.md)
+       3. Return formatted summary to orchestrator
+  → Orchestrator performs MCP saves directly:
+       1. claude-mem save (if available via ToolSearch)
+       (episodic-memory auto-indexes after session — no action needed)
+  → Orchestrator confirms to user
+```
+
+### Responsibility Split
+
+MCP tools (claude-mem, episodic-memory) are **orchestrator-scoped** and not inherited by subagents. Therefore:
+
+| Responsibility | Owner | Reason |
+|----------------|-------|--------|
+| Session summary collection | sys-memory-keeper | Domain expertise in memory formatting |
+| Native auto-memory (MEMORY.md) | sys-memory-keeper | Has Write access to memory directory |
+| claude-mem MCP save | Orchestrator | MCP tools only available at orchestrator level |
+| episodic-memory | Automatic | Conversations are auto-indexed after session ends — no manual action needed |
+
+### Dual-System Save
+
+| System | Owner | Tool | Action | Required |
+|--------|-------|------|--------|----------|
+| Native auto-memory | sys-memory-keeper | Write | Update MEMORY.md with session learnings | Yes |
+| claude-mem | Orchestrator | `mcp__plugin_claude-mem_mcp-search__save_memory` | Save session summary with project, tasks, decisions | No (best-effort) |
+| episodic-memory | Automatic | (auto-indexed) | No action needed — conversations are indexed automatically after session ends | N/A |
+
+### Session-End Self-Check (MANDATORY)
+
+```
+╔══════════════════════════════════════════════════════════════════╗
+║  BEFORE CONFIRMING SESSION-END TO USER:                          ║
+║                                                                   ║
+║  1. Did sys-memory-keeper update MEMORY.md?                      ║
+║     YES → Continue                                               ║
+║     NO  → Delegate to sys-memory-keeper first                    ║
+║                                                                   ║
+║  2. Did I attempt claude-mem save?                               ║
+║     YES → Continue (even if it failed)                           ║
+║     NO  → ToolSearch + save now                                  ║
+║                                                                   ║
+║  Note: episodic-memory auto-indexes conversations after session  ║
+║  ends. No manual action needed — do NOT search as "verification" ║
+║                                                                   ║
+║  BOTH steps must be completed before confirming to user.         ║
+║  "Attempted" means called the tool — failure is OK, skipping     ║
+║  is NOT.                                                          ║
+╚══════════════════════════════════════════════════════════════════╝
+```
+
+### Failure Policy
+
+- MCP saves are **non-blocking**: memory failure MUST NOT prevent session from ending
+- If claude-mem unavailable: skip, log warning
+- episodic-memory: no action needed (auto-indexed after session)

package/templates/.claude/rules/SHOULD-ontology-rag-routing.md

@@ -0,0 +1,49 @@
+# [SHOULD] Ontology-RAG Assisted Routing
+
+> **Priority**: SHOULD | **ID**: R019
+
+## Core Rule
+
+Routing skills SHOULD use ontology-RAG's `get_agent_for_task` MCP tool to enrich agent selection with contextual skill suggestions. Ontology-RAG is an enrichment layer — it does NOT replace static routing.
+
+## Integration Pattern
+
+After static routing selects an agent, call `get_agent_for_task(query)` and extract `suggested_skills` from the response. Inject these into the spawned agent's prompt as contextual hints.
+
+```
+Static routing → agent selected
+  ↓
+get_agent_for_task(original_query)
+  ↓
+Extract suggested_skills
+  ↓
+Prepend to spawned agent prompt:
+  "Ontology context suggests these skills may be relevant: {suggested_skills}"
+```
+
+## Failure Handling
+
+| Scenario | Action |
+|----------|--------|
+| MCP server unavailable | Skip silently, proceed with unmodified prompt |
+| get_agent_for_task returns empty | Proceed with unmodified prompt |
+| Response parsing error | Skip silently, log warning |
+
+**MCP failure MUST NOT block or delay routing.** Ontology-RAG is advisory only.
+
+## Scope
+
+| Applies to | Details |
+|------------|---------|
+| secretary-routing | Enriches manager agent selection |
+| dev-lead-routing | Enriches language/framework expert selection |
+| de-lead-routing | Enriches data engineering expert selection |
+| qa-lead-routing | Enriches QA workflow routing |
+
+## Interaction with Other Rules
+
+| Rule | Interaction |
+|------|-------------|
+| R010 | Orchestrator calls MCP tool; subagent receives enriched prompt |
+| R015 | Post-fix confidence (0.30-0.40) remains below R015's 70% threshold; display behavior unchanged |
+| R009 | Ontology-RAG call adds ~300 tokens; no parallelism impact |

package/templates/.claude/rules/SHOULD-wiki-sync.md

@@ -0,0 +1,73 @@
+# [SHOULD] Wiki Sync Rules
+
+> **Priority**: SHOULD | **ID**: R022
+
+## Core Rule
+
+When agents, skills, rules, or guides are created or modified, corresponding wiki pages SHOULD be updated to keep the knowledge base current. The wiki is the project's compiled knowledge — stale wikis degrade team and LLM onboarding speed.
+
+## When to Sync
+
+| Change Type | Wiki Action |
+|-------------|-------------|
+| New agent created | Create wiki/agents/{name}.md |
+| Agent modified | Update wiki/agents/{name}.md |
+| New skill created | Create wiki/skills/{name}.md |
+| Skill modified | Update wiki/skills/{name}.md |
+| Rule created/modified | Update wiki/rules/r{nnn}.md |
+| Guide created/modified | Update wiki/guides/{name}.md |
+| Architecture change | Update wiki/architecture/ pages |
+| Multiple changes | Run `/omcodex:wiki` for full update |
+
+## How to Sync
+
+| Method | When |
+|--------|------|
+| `/omcodex:wiki ingest <path>` | Single file/directory changed |
+| `/omcodex:wiki` | Multiple files changed or periodic refresh |
+| `/omcodex:wiki lint` | After major structural changes |
+| Automatic (CI) | `.github/workflows/wiki-sync.yml` checks on PR |
+
+## Delegation — All wiki writes via wiki-curator agent (R010). See workflow via Read tool.
+
+<!-- DETAIL: Delegation
+All wiki writes MUST go through the `wiki-curator` agent (R010). The orchestrator reads wiki pages freely but never writes them directly.
+
+```
+Orchestrator
+├── Detects source change
+├── Delegates to wiki-curator
+│   ├── Reads source file
+│   ├── Creates/updates wiki page
+│   ├── Updates cross-references
+│   └── Updates index.md
+└── Verifies via wiki lint
+```
+-->
+
+## Integration — Interacts with R010, R017, R020, R006, R021. See table via Read tool.
+
+<!-- DETAIL: Integration
+| Rule | Interaction |
+|------|-------------|
+| R010 | Wiki writes delegated to wiki-curator agent |
+| R017 | Wiki sync added to sauron verification Phase 3 |
+| R020 | Wiki-dependent tasks verify wiki is current before [Done] |
+| R006 | Wiki pages follow same separation of concerns as source |
+| R021 | SHOULD priority — advisory enforcement, CI check |
+-->
+
+## CI Enforcement
+
+`.github/workflows/wiki-sync.yml` checks for missing wiki pages on every PR. Missing pages cause CI failure with guidance to run `/omcodex:wiki`.
+
+## Self-Check — 3 checks: wiki pages updated, index refreshed, lint passed. See details via Read tool.
+
+<!-- DETAIL: Self-Check
+Before completing a session that modified agents/skills/rules/guides:
+1. Were wiki pages updated for all changes?
+2. Was index.md refreshed?
+3. Did wiki lint pass?
+
+If any NO → run `/omcodex:wiki ingest` for affected paths.
+-->

package/templates/.claude/rules/index.yaml

@@ -0,0 +1,141 @@
+# Global Rules
+# Priority: MUST > SHOULD > MAY
+
+rules:
+  # MUST - Never violate
+  - id: R000
+    name: language-policy
+    title: Language & Delegation Policy
+    path: ./MUST-language-policy.md
+    priority: MUST
+    scope: all
+
+  - id: R001
+    name: safety
+    title: Safety Rules
+    path: ./MUST-safety.md
+    priority: MUST
+    scope: all
+
+  - id: R002
+    name: permissions
+    title: Permission Rules
+    path: ./MUST-permissions.md
+    priority: MUST
+    scope: all
+
+  - id: R007
+    name: agent-identification
+    title: Agent Identification Rules
+    path: ./MUST-agent-identification.md
+    priority: MUST
+    scope: all
+
+  - id: R008
+    name: tool-identification
+    title: Tool Usage Identification Rules
+    path: ./MUST-tool-identification.md
+    priority: MUST
+    scope: all
+
+  - id: R009
+    name: parallel-execution
+    title: Parallel Execution Rules
+    path: ./MUST-parallel-execution.md
+    priority: MUST
+    scope: all
+
+  # Agent Design - MUST
+  - id: R006
+    name: agent-design
+    title: Agent Design Rules
+    path: ./MUST-agent-design.md
+    priority: MUST
+    scope: agents
+
+  # Orchestrator Coordination - MUST
+  - id: R010
+    name: orchestrator-coordination
+    title: Orchestrator Coordination Rules
+    path: ./MUST-orchestrator-coordination.md
+    priority: MUST
+    scope: agents
+
+  # Intent Transparency - MUST
+  - id: R015
+    name: intent-transparency
+    title: Intent Transparency Rules
+    path: ./MUST-intent-transparency.md
+    priority: MUST
+    scope: orchestrator
+
+  # Continuous Improvement - MUST
+  - id: R016
+    name: continuous-improvement
+    title: Continuous Improvement Rules
+    path: ./MUST-continuous-improvement.md
+    priority: MUST
+    scope: all
+
+  # Sync Verification - MUST
+  - id: R017
+    name: sync-verification
+    title: Sync Verification Rules
+    path: ./MUST-sync-verification.md
+    priority: MUST
+    scope: all
+
+  # SHOULD - Strongly recommended
+  - id: R003
+    name: interaction
+    title: Interaction Rules
+    path: ./SHOULD-interaction.md
+    priority: SHOULD
+    scope: all
+
+  - id: R004
+    name: error-handling
+    title: Error Handling Rules
+    path: ./SHOULD-error-handling.md
+    priority: SHOULD
+    scope: all
+
+  # Memory Integration - SHOULD
+  - id: R011
+    name: memory-integration
+    title: Memory Integration Rules
+    path: ./SHOULD-memory-integration.md
+    priority: SHOULD
+    scope: all
+
+  # HUD Statusline - SHOULD
+  - id: R012
+    name: hud-statusline
+    title: HUD Statusline Rules
+    path: ./SHOULD-hud-statusline.md
+    priority: SHOULD
+    scope: all
+
+  # Ecomode - SHOULD
+  - id: R013
+    name: ecomode
+    title: Ecomode Rules
+    path: ./SHOULD-ecomode.md
+    priority: SHOULD
+    scope: all
+
+  # Agent Teams - MUST (Conditional)
+  - id: R018
+    name: agent-teams
+    title: Agent Teams Rules (Conditional)
+    path: ./MUST-agent-teams.md
+    priority: MUST
+    scope: orchestrator
+
+  # MAY - Optional
+  - id: R005
+    name: optimization
+    title: Optimization Guide
+    path: ./MAY-optimization.md
+    priority: MAY
+    scope: all

package/templates/.claude/schemas/tool-inputs.json

@@ -0,0 +1,62 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "description": "Tool input validation schemas for PreToolUse hook",
+  "definitions": {
+    "Write": {
+      "type": "object",
+      "required": ["file_path", "content"],
+      "properties": {
+        "file_path": {
+          "type": "string",
+          "minLength": 1,
+          "description": "Absolute file path to write"
+        },
+        "content": {
+          "type": "string",
+          "description": "File content to write"
+        }
+      }
+    },
+    "Edit": {
+      "type": "object",
+      "required": ["file_path", "old_string", "new_string"],
+      "properties": {
+        "file_path": {
+          "type": "string",
+          "minLength": 1,
+          "description": "Absolute file path to edit"
+        },
+        "old_string": {
+          "type": "string",
+          "minLength": 1,
+          "description": "Text to find and replace"
+        },
+        "new_string": {
+          "type": "string",
+          "description": "Replacement text"
+        }
+      },
+      "additionalValidation": {
+        "rule": "old_string !== new_string",
+        "message": "old_string and new_string must be different"
+      }
+    },
+    "Bash": {
+      "type": "object",
+      "required": ["command"],
+      "properties": {
+        "command": {
+          "type": "string",
+          "minLength": 1,
+          "description": "Shell command to execute"
+        }
+      },
+      "dangerousPatterns": [
+        { "pattern": "rm\\s+-rf\\s+/", "message": "Dangerous recursive delete from root" },
+        { "pattern": "sudo\\s+", "message": "Elevated privilege command detected" },
+        { "pattern": "> /dev/sd", "message": "Direct disk write detected" },
+        { "pattern": "mkfs\\.", "message": "Filesystem format command detected" }
+      ]
+    }
+  }
+}