diff --git a/CHANGELOG.md b/CHANGELOG.md index b59a4e4..53fc0a6 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -87,6 +87,36 @@ Per-agent model selection using Claude Code's now-supported `model` frontmatter - Added missing frontmatter to 13 agents across pr-review, viz-platform, contract-validator, clarity-assist, git-flow, doc-guardian, code-sentinel, cmdb-assistant, and data-platform - All 25 agents now have consistent `name`, `description`, and `model` fields +### Changed + +#### Agent Frontmatter Hardening v3 + +Comprehensive agent-level configuration using Claude Code's supported frontmatter fields. + +**permissionMode added to all 25 agents:** +- `bypassPermissions` (1): Executor — full autonomy with code-sentinel + Code Reviewer safety nets +- `acceptEdits` (7): Orchestrator, Data Ingestion, Theme Setup, Refactor Advisor, Doc Analyzer, Git Assistant, Maintainer +- `default` (7): Planner, Code Reviewer, Data Advisor, Layout Builder, Full Validation, Clarity Coach, CMDB Assistant +- `plan` (10): All pr-review agents (5), Data Analysis, Design Reviewer, Component Check, Agent Check, Security Reviewer (code-sentinel) + +**disallowedTools added to 12 agents:** +- All `plan`-mode agents (10) + Code Reviewer + Clarity Coach receive `disallowedTools: Write, Edit, MultiEdit` +- Enforces read-only contracts at platform level (defense-in-depth with `permissionMode`) + +**Model promotions:** +- Planner: `sonnet` → `opus` (architectural reasoning benefits from deeper analysis) +- Code Reviewer: `sonnet` → `opus` (quality gate benefits from thorough review) + +**skills frontmatter on 3 agents:** +- Executor: 7 safety-critical skills auto-injected (branch-security, runaway-detection, etc.) +- Code Reviewer: 4 review skills auto-injected +- Maintainer: 2 config skills auto-injected +- Body text `## Skills to Load` removed for these agents to avoid duplication + +**Documentation:** +- `CLAUDE.md` and `docs/CONFIGURATION.md` updated with complete agent configuration matrix +- New subsections: permissionMode Guide, disallowedTools Guide, skills Frontmatter Guide + --- ## [5.8.0] - 2026-02-02 diff --git a/CLAUDE.md b/CLAUDE.md index dda92d0..687ef6b 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -271,39 +271,60 @@ leo-claude-mktplace/ | **Executor** | Implementation-focused | Code implementation, branch management, MR creation | | **Code Reviewer** | Thorough, practical | Pre-close quality review, security scan, test verification | -### Agent Model Selection +### Agent Frontmatter Configuration -Agents specify their model in frontmatter using Claude Code's `model` field. Supported values: `sonnet` (default), `opus`, `haiku`, `inherit`. +Agents specify their configuration in frontmatter using Claude Code's supported fields. Reference: https://code.claude.com/docs/en/sub-agents -| Plugin | Agent | Model | Rationale | -|--------|-------|-------|-----------| -| projman | Planner | sonnet | Architectural analysis, sprint planning | -| projman | Orchestrator | sonnet | Coordination and tool dispatch | -| projman | Executor | sonnet | Code generation and implementation | -| projman | Code Reviewer | sonnet | Quality gate, pattern detection | -| pr-review | Coordinator | sonnet | Orchestrates sub-agents, aggregates findings | -| pr-review | Security Reviewer | sonnet | Security analysis | -| pr-review | Performance Analyst | sonnet | Performance pattern detection | -| pr-review | Maintainability Auditor | haiku | Pattern matching (complexity, duplication) | -| pr-review | Test Validator | haiku | Coverage gap detection | -| data-platform | Data Advisor | sonnet | Schema validation, dbt orchestration | -| data-platform | Data Analysis | sonnet | Data exploration and profiling | -| data-platform | Data Ingestion | haiku | Data loading operations | -| viz-platform | Design Reviewer | sonnet | DMC validation + accessibility | -| viz-platform | Layout Builder | sonnet | Dashboard design guidance | -| viz-platform | Component Check | haiku | Quick component validation | -| viz-platform | Theme Setup | haiku | Theme configuration | -| contract-validator | Agent Check | haiku | Reference checking | -| contract-validator | Full Validation | sonnet | Marketplace sweep | -| code-sentinel | Security Reviewer | sonnet | Security analysis | -| code-sentinel | Refactor Advisor | sonnet | Code refactoring advice | -| doc-guardian | Doc Analyzer | sonnet | Documentation drift detection | -| clarity-assist | Clarity Coach | sonnet | Conversational coaching | -| git-flow | Git Assistant | haiku | Git operations | -| claude-config-maintainer | Maintainer | sonnet | CLAUDE.md optimization | -| cmdb-assistant | CMDB Assistant | sonnet | NetBox operations | +**Supported frontmatter fields:** -Override by editing the `model:` field in `plugins/{plugin}/agents/{agent}.md`. +| Field | Required | Default | Description | +|-------|----------|---------|-------------| +| `name` | Yes | — | Unique identifier, lowercase + hyphens | +| `description` | Yes | — | When Claude should delegate to this subagent | +| `model` | No | `inherit` | `sonnet`, `opus`, `haiku`, or `inherit` | +| `permissionMode` | No | `default` | Controls permission prompts: `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan` | +| `disallowedTools` | No | none | Comma-separated tools to remove from agent's toolset | +| `skills` | No | none | Comma-separated skills auto-injected into context at startup | +| `hooks` | No | none | Lifecycle hooks scoped to this subagent | + +**Complete agent matrix:** + +| Plugin | Agent | `model` | `permissionMode` | `disallowedTools` | `skills` | +|--------|-------|---------|-------------------|--------------------|----------| +| projman | planner | opus | default | — | body text (14) | +| projman | orchestrator | sonnet | acceptEdits | — | body text (12) | +| projman | executor | sonnet | bypassPermissions | — | frontmatter (7) | +| projman | code-reviewer | opus | default | Write, Edit, MultiEdit | frontmatter (4) | +| pr-review | coordinator | sonnet | plan | Write, Edit, MultiEdit | — | +| pr-review | security-reviewer | sonnet | plan | Write, Edit, MultiEdit | — | +| pr-review | performance-analyst | sonnet | plan | Write, Edit, MultiEdit | — | +| pr-review | maintainability-auditor | haiku | plan | Write, Edit, MultiEdit | — | +| pr-review | test-validator | haiku | plan | Write, Edit, MultiEdit | — | +| data-platform | data-advisor | sonnet | default | — | — | +| data-platform | data-analysis | sonnet | plan | Write, Edit, MultiEdit | — | +| data-platform | data-ingestion | haiku | acceptEdits | — | — | +| viz-platform | design-reviewer | sonnet | plan | Write, Edit, MultiEdit | — | +| viz-platform | layout-builder | sonnet | default | — | — | +| viz-platform | component-check | haiku | plan | Write, Edit, MultiEdit | — | +| viz-platform | theme-setup | haiku | acceptEdits | — | — | +| contract-validator | full-validation | sonnet | default | — | — | +| contract-validator | agent-check | haiku | plan | Write, Edit, MultiEdit | — | +| code-sentinel | security-reviewer | sonnet | plan | Write, Edit, MultiEdit | — | +| code-sentinel | refactor-advisor | sonnet | acceptEdits | — | — | +| doc-guardian | doc-analyzer | sonnet | acceptEdits | — | — | +| clarity-assist | clarity-coach | sonnet | default | Write, Edit, MultiEdit | — | +| git-flow | git-assistant | haiku | acceptEdits | — | — | +| claude-config-maintainer | maintainer | sonnet | acceptEdits | — | frontmatter (2) | +| cmdb-assistant | cmdb-assistant | sonnet | default | — | — | + +**Design principles:** +- `bypassPermissions` is granted to exactly ONE agent (Executor) which has code-sentinel PreToolUse hook + Code Reviewer downstream as safety nets. +- `plan` mode is assigned to all pure analysis agents (pr-review, read-only validators). +- `disallowedTools: Write, Edit, MultiEdit` provides defense-in-depth on agents that should never write files. +- `skills` frontmatter is used for agents with ≤7 skills where guaranteed loading is safety-critical. Agents with 8+ skills use body text `## Skills to Load` for selective loading. +- `hooks` (agent-scoped) is reserved for future use (v6.0+). + +Override any field by editing the agent's `.md` file in `plugins/{plugin}/agents/`. ### MCP Server Tools (Gitea) diff --git a/docs/CONFIGURATION.md b/docs/CONFIGURATION.md index ed6e6b6..4664342 100644 --- a/docs/CONFIGURATION.md +++ b/docs/CONFIGURATION.md @@ -496,60 +496,100 @@ Not all plugins have MCP servers. The install script handles this automatically: --- -## Agent Model Selection +## Agent Frontmatter Configuration -Marketplace agents specify their preferred model using Claude Code's `model` frontmatter field. This allows cost/performance optimization per agent. +Agents specify their configuration in frontmatter using Claude Code's supported fields. Reference: https://code.claude.com/docs/en/sub-agents -### Supported Values +### Supported Frontmatter Fields -| Value | Description | -|-------|-------------| -| `sonnet` | Default. Balanced performance and cost. | -| `opus` | Higher reasoning depth. Use for complex analysis. | -| `haiku` | Faster, lower cost. Use for mechanical tasks. | -| `inherit` | Use session's current model setting. | +| Field | Required | Default | Description | +|-------|----------|---------|-------------| +| `name` | Yes | — | Unique identifier, lowercase + hyphens | +| `description` | Yes | — | When Claude should delegate to this subagent | +| `model` | No | `inherit` | `sonnet`, `opus`, `haiku`, or `inherit` | +| `permissionMode` | No | `default` | Controls permission prompts: `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan` | +| `disallowedTools` | No | none | Comma-separated tools to remove from agent's toolset | +| `skills` | No | none | Comma-separated skills auto-injected into context at startup | +| `hooks` | No | none | Lifecycle hooks scoped to this subagent | -### How It Works +### Complete Agent Matrix -Each agent in `plugins/{plugin}/agents/{agent}.md` has frontmatter like: +| Plugin | Agent | `model` | `permissionMode` | `disallowedTools` | `skills` | +|--------|-------|---------|-------------------|--------------------|----------| +| projman | planner | opus | default | — | body text (14) | +| projman | orchestrator | sonnet | acceptEdits | — | body text (12) | +| projman | executor | sonnet | bypassPermissions | — | frontmatter (7) | +| projman | code-reviewer | opus | default | Write, Edit, MultiEdit | frontmatter (4) | +| pr-review | coordinator | sonnet | plan | Write, Edit, MultiEdit | — | +| pr-review | security-reviewer | sonnet | plan | Write, Edit, MultiEdit | — | +| pr-review | performance-analyst | sonnet | plan | Write, Edit, MultiEdit | — | +| pr-review | maintainability-auditor | haiku | plan | Write, Edit, MultiEdit | — | +| pr-review | test-validator | haiku | plan | Write, Edit, MultiEdit | — | +| data-platform | data-advisor | sonnet | default | — | — | +| data-platform | data-analysis | sonnet | plan | Write, Edit, MultiEdit | — | +| data-platform | data-ingestion | haiku | acceptEdits | — | — | +| viz-platform | design-reviewer | sonnet | plan | Write, Edit, MultiEdit | — | +| viz-platform | layout-builder | sonnet | default | — | — | +| viz-platform | component-check | haiku | plan | Write, Edit, MultiEdit | — | +| viz-platform | theme-setup | haiku | acceptEdits | — | — | +| contract-validator | full-validation | sonnet | default | — | — | +| contract-validator | agent-check | haiku | plan | Write, Edit, MultiEdit | — | +| code-sentinel | security-reviewer | sonnet | plan | Write, Edit, MultiEdit | — | +| code-sentinel | refactor-advisor | sonnet | acceptEdits | — | — | +| doc-guardian | doc-analyzer | sonnet | acceptEdits | — | — | +| clarity-assist | clarity-coach | sonnet | default | Write, Edit, MultiEdit | — | +| git-flow | git-assistant | haiku | acceptEdits | — | — | +| claude-config-maintainer | maintainer | sonnet | acceptEdits | — | frontmatter (2) | +| cmdb-assistant | cmdb-assistant | sonnet | default | — | — | -```yaml ---- -name: planner -description: Sprint planning agent - thoughtful architecture analysis -model: sonnet ---- -``` +### Design Principles -Claude Code reads this field when invoking the agent as a subagent. +- `bypassPermissions` is granted to exactly ONE agent (Executor) which has code-sentinel PreToolUse hook + Code Reviewer downstream as safety nets. +- `plan` mode is assigned to all pure analysis agents (pr-review, read-only validators). +- `disallowedTools: Write, Edit, MultiEdit` provides defense-in-depth on agents that should never write files. +- `skills` frontmatter is used for agents with ≤7 skills where guaranteed loading is safety-critical. Agents with 8+ skills use body text `## Skills to Load` for selective loading. +- `hooks` (agent-scoped) is reserved for future use (v6.0+). -### Model Assignments +Override any field by editing the agent's `.md` file in `plugins/{plugin}/agents/`. -Agents are assigned models based on their task complexity: +### permissionMode Guide -| Model | Agents | Rationale | -|-------|--------|-----------| -| **sonnet** | Planner, Orchestrator, Executor, Code Reviewer, Coordinator, Security Reviewers, Performance Analyst, Data Advisor, Data Analysis, Design Reviewer, Layout Builder, Full Validation, Doc Analyzer, Clarity Coach, Maintainer, CMDB Assistant, Refactor Advisor | Standard reasoning, tool orchestration, code generation | -| **haiku** | Maintainability Auditor, Test Validator, Component Check, Theme Setup, Agent Check, Data Ingestion, Git Assistant | Pattern matching, quick validation, mechanical tasks | +| Value | Prompts for file ops? | Prompts for Bash? | Prompts for MCP? | Use when | +|-------|-----------------------|-------------------|-------------------|----------| +| `default` | Yes | Yes | No (MCP bypasses permissions) | You want full visibility | +| `acceptEdits` | No | Yes | No | Core job is file read/write, Bash visibility useful | +| `dontAsk` | No | No (most) | No | Even Bash prompts are friction | +| `bypassPermissions` | No | No | No | Agent has downstream safety layers | +| `plan` | N/A (read-only) | N/A (read-only) | No | Pure analysis, no modifications | -### Overriding Model Selection +### disallowedTools Guide -**Per-agent override:** Edit the `model:` field in the agent file: +Use `disallowedTools` to remove specific tools from an agent's toolset. This is a blacklist — the agent inherits all tools from the main thread, then the listed tools are removed. -```bash -# Change executor to use opus for heavy implementation work -nano plugins/projman/agents/executor.md -# Change model: sonnet to model: opus -``` +Prefer `disallowedTools` over `tools` (whitelist) because: +- New MCP servers are automatically available without updating every agent. +- Less configuration to maintain. +- Easier to audit — you only list what's blocked. -**Session-level:** Users on Opus subscription can change the agent's model to `inherit` to use whatever model the session is using. +Common patterns: +- `disallowedTools: Write, Edit, MultiEdit` — read-only agent, cannot modify files. +- `disallowedTools: Bash` — no shell access (rare, most agents need at least read-only Bash). -### Best Practices +### skills Frontmatter Guide -1. **Default to sonnet** - Good balance for most tasks -2. **Use haiku for speed-sensitive agents** - Sub-agents dispatched in parallel, read-only tasks -3. **Reserve opus for heavy analysis** - Only when sonnet's reasoning isn't sufficient -4. **Use inherit sparingly** - Only when you want session-level control +The `skills` field auto-injects skill file contents into the agent's context window at startup. The agent does NOT need to read the files — they are already present. + +**When to use frontmatter `skills`:** +- Agent has ≤7 skills. +- Skills are safety-critical (e.g., `branch-security`, `runaway-detection`). +- You need guaranteed loading — no risk of the agent skipping a skill. + +**When to keep body text `## Skills to Load`:** +- Agent has 8+ skills (context window cost too high for full injection). +- Skills are situational — not all needed for every invocation. +- Agent benefits from selective loading based on the specific task. + +Skill names in frontmatter are resolved relative to the plugin's `skills/` directory. Use the filename without the `.md` extension. --- diff --git a/plugins/clarity-assist/agents/clarity-coach.md b/plugins/clarity-assist/agents/clarity-coach.md index 1c2c62f..d9100d6 100644 --- a/plugins/clarity-assist/agents/clarity-coach.md +++ b/plugins/clarity-assist/agents/clarity-coach.md @@ -2,6 +2,8 @@ name: clarity-coach description: Patient, structured coach helping users articulate requirements clearly. Uses neurodivergent-friendly communication patterns. model: sonnet +permissionMode: default +disallowedTools: Write, Edit, MultiEdit --- # Clarity Coach Agent diff --git a/plugins/claude-config-maintainer/agents/maintainer.md b/plugins/claude-config-maintainer/agents/maintainer.md index 550273a..b27159a 100644 --- a/plugins/claude-config-maintainer/agents/maintainer.md +++ b/plugins/claude-config-maintainer/agents/maintainer.md @@ -2,6 +2,8 @@ name: maintainer description: CLAUDE.md optimization and maintenance agent model: sonnet +permissionMode: acceptEdits +skills: visual-header, settings-optimization --- # CLAUDE.md Maintainer Agent diff --git a/plugins/cmdb-assistant/agents/cmdb-assistant.md b/plugins/cmdb-assistant/agents/cmdb-assistant.md index 791a322..725d42e 100644 --- a/plugins/cmdb-assistant/agents/cmdb-assistant.md +++ b/plugins/cmdb-assistant/agents/cmdb-assistant.md @@ -2,6 +2,7 @@ name: cmdb-assistant description: Infrastructure management assistant specialized in NetBox CMDB operations. Use for device management, IP addressing, and infrastructure queries. model: sonnet +permissionMode: default --- # CMDB Assistant Agent diff --git a/plugins/code-sentinel/agents/refactor-advisor.md b/plugins/code-sentinel/agents/refactor-advisor.md index ff34167..d40991e 100644 --- a/plugins/code-sentinel/agents/refactor-advisor.md +++ b/plugins/code-sentinel/agents/refactor-advisor.md @@ -2,6 +2,7 @@ name: refactor-advisor description: Code structure and refactoring specialist. Use when analyzing code quality, design patterns, or planning refactoring work. model: sonnet +permissionMode: acceptEdits --- # Refactor Advisor Agent diff --git a/plugins/code-sentinel/agents/security-reviewer.md b/plugins/code-sentinel/agents/security-reviewer.md index 3f1dee7..2891bad 100644 --- a/plugins/code-sentinel/agents/security-reviewer.md +++ b/plugins/code-sentinel/agents/security-reviewer.md @@ -2,6 +2,8 @@ name: security-reviewer description: Security-focused code review agent model: sonnet +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Security Reviewer Agent diff --git a/plugins/contract-validator/agents/agent-check.md b/plugins/contract-validator/agents/agent-check.md index 39c5a51..8715f3b 100644 --- a/plugins/contract-validator/agents/agent-check.md +++ b/plugins/contract-validator/agents/agent-check.md @@ -2,6 +2,8 @@ name: agent-check description: Agent definition validator for quick verification model: haiku +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Agent Check Agent diff --git a/plugins/contract-validator/agents/full-validation.md b/plugins/contract-validator/agents/full-validation.md index 5fc9767..01f1272 100644 --- a/plugins/contract-validator/agents/full-validation.md +++ b/plugins/contract-validator/agents/full-validation.md @@ -2,6 +2,7 @@ name: full-validation description: Contract validation specialist for comprehensive cross-plugin compatibility validation of the entire marketplace. model: sonnet +permissionMode: default --- # Full Validation Agent diff --git a/plugins/data-platform/agents/data-advisor.md b/plugins/data-platform/agents/data-advisor.md index d54ea1e..1d00581 100644 --- a/plugins/data-platform/agents/data-advisor.md +++ b/plugins/data-platform/agents/data-advisor.md @@ -2,6 +2,7 @@ name: data-advisor description: Reviews code for data integrity, schema validity, and dbt compliance using data-platform MCP tools. Use when validating database operations or data pipelines. model: sonnet +permissionMode: default --- # Data Advisor Agent diff --git a/plugins/data-platform/agents/data-analysis.md b/plugins/data-platform/agents/data-analysis.md index 3f47a35..1b48019 100644 --- a/plugins/data-platform/agents/data-analysis.md +++ b/plugins/data-platform/agents/data-analysis.md @@ -2,6 +2,8 @@ name: data-analysis description: Data analysis specialist for exploration and profiling model: sonnet +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Data Analysis Agent diff --git a/plugins/data-platform/agents/data-ingestion.md b/plugins/data-platform/agents/data-ingestion.md index 1a88768..5a7fef6 100644 --- a/plugins/data-platform/agents/data-ingestion.md +++ b/plugins/data-platform/agents/data-ingestion.md @@ -2,6 +2,7 @@ name: data-ingestion description: Data ingestion specialist for loading, transforming, and preparing data for analysis. model: haiku +permissionMode: acceptEdits --- # Data Ingestion Agent diff --git a/plugins/doc-guardian/agents/doc-analyzer.md b/plugins/doc-guardian/agents/doc-analyzer.md index 2599360..a27645c 100644 --- a/plugins/doc-guardian/agents/doc-analyzer.md +++ b/plugins/doc-guardian/agents/doc-analyzer.md @@ -2,6 +2,7 @@ name: doc-analyzer description: Specialized agent for documentation analysis and drift detection. Use when detecting or fixing discrepancies between code and documentation. model: sonnet +permissionMode: acceptEdits --- # Documentation Analyzer Agent diff --git a/plugins/git-flow/agents/git-assistant.md b/plugins/git-flow/agents/git-assistant.md index 65c02e2..0153dfd 100644 --- a/plugins/git-flow/agents/git-assistant.md +++ b/plugins/git-flow/agents/git-assistant.md @@ -2,6 +2,7 @@ name: git-assistant description: Git workflow assistant for complex git operations, conflict resolution, and repository history management. model: haiku +permissionMode: acceptEdits --- # Git Assistant Agent diff --git a/plugins/pr-review/agents/coordinator.md b/plugins/pr-review/agents/coordinator.md index ec3ee8d..f0d2321 100644 --- a/plugins/pr-review/agents/coordinator.md +++ b/plugins/pr-review/agents/coordinator.md @@ -2,6 +2,8 @@ name: coordinator description: Review coordinator that orchestrates the multi-agent PR review process. Dispatches to specialized reviewers, aggregates findings, and produces the final review report. Use proactively after code changes. model: sonnet +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Coordinator Agent diff --git a/plugins/pr-review/agents/maintainability-auditor.md b/plugins/pr-review/agents/maintainability-auditor.md index c605a82..36b7748 100644 --- a/plugins/pr-review/agents/maintainability-auditor.md +++ b/plugins/pr-review/agents/maintainability-auditor.md @@ -2,6 +2,8 @@ name: maintainability-auditor description: Identifies code complexity, duplication, naming issues, and architecture concerns in PR changes. model: haiku +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Maintainability Auditor Agent diff --git a/plugins/pr-review/agents/performance-analyst.md b/plugins/pr-review/agents/performance-analyst.md index 17af156..90b269f 100644 --- a/plugins/pr-review/agents/performance-analyst.md +++ b/plugins/pr-review/agents/performance-analyst.md @@ -2,6 +2,8 @@ name: performance-analyst description: Performance-focused code reviewer that identifies performance issues, inefficiencies, and optimization opportunities. model: sonnet +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Performance Analyst Agent diff --git a/plugins/pr-review/agents/security-reviewer.md b/plugins/pr-review/agents/security-reviewer.md index ba3b76f..c8d5796 100644 --- a/plugins/pr-review/agents/security-reviewer.md +++ b/plugins/pr-review/agents/security-reviewer.md @@ -2,6 +2,8 @@ name: security-reviewer description: Security-focused code reviewer for PR analysis model: sonnet +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Security Reviewer Agent diff --git a/plugins/pr-review/agents/test-validator.md b/plugins/pr-review/agents/test-validator.md index eee25ea..c79de21 100644 --- a/plugins/pr-review/agents/test-validator.md +++ b/plugins/pr-review/agents/test-validator.md @@ -2,6 +2,8 @@ name: test-validator description: Test quality reviewer that validates test coverage, test quality, and testing practices in PR changes. model: haiku +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Test Validator Agent diff --git a/plugins/projman/agents/code-reviewer.md b/plugins/projman/agents/code-reviewer.md index 8209808..14fa273 100644 --- a/plugins/projman/agents/code-reviewer.md +++ b/plugins/projman/agents/code-reviewer.md @@ -1,20 +1,16 @@ --- name: code-reviewer description: Pre-sprint code quality review agent -model: sonnet +model: opus +permissionMode: default +disallowedTools: Write, Edit, MultiEdit +skills: review-checklist, test-standards, sprint-lifecycle, visual-output --- # Code Reviewer Agent You are the **Code Reviewer Agent** - a thorough, practical reviewer who ensures code quality before sprint close. -## Skills to Load - -- skills/review-checklist.md -- skills/test-standards.md -- skills/sprint-lifecycle.md -- skills/visual-output.md - ## Your Personality **Thorough but Practical:** diff --git a/plugins/projman/agents/executor.md b/plugins/projman/agents/executor.md index 5813712..83831ae 100644 --- a/plugins/projman/agents/executor.md +++ b/plugins/projman/agents/executor.md @@ -2,22 +2,14 @@ name: executor description: Implementation executor agent - precise implementation guidance and code quality model: sonnet +permissionMode: bypassPermissions +skills: mcp-tools-reference, branch-security, git-workflow, progress-tracking, runaway-detection, lessons-learned, visual-output --- # Implementation Executor Agent You are the **Executor Agent** - an implementation-focused specialist who writes clean code and ensures quality. -## Skills to Load - -- skills/mcp-tools-reference.md -- skills/branch-security.md -- skills/git-workflow.md -- skills/progress-tracking.md -- skills/runaway-detection.md -- skills/lessons-learned.md -- skills/visual-output.md - ## Your Personality **Implementation-Focused:** diff --git a/plugins/projman/agents/orchestrator.md b/plugins/projman/agents/orchestrator.md index 9651060..3f77d89 100644 --- a/plugins/projman/agents/orchestrator.md +++ b/plugins/projman/agents/orchestrator.md @@ -2,6 +2,7 @@ name: orchestrator description: Sprint orchestration agent - coordinates execution and tracks progress model: sonnet +permissionMode: acceptEdits --- # Sprint Orchestration Agent diff --git a/plugins/projman/agents/planner.md b/plugins/projman/agents/planner.md index 65a3820..202e7d4 100644 --- a/plugins/projman/agents/planner.md +++ b/plugins/projman/agents/planner.md @@ -1,7 +1,8 @@ --- name: planner description: Sprint planning agent - thoughtful architecture analysis and issue creation -model: sonnet +model: opus +permissionMode: default --- # Sprint Planning Agent diff --git a/plugins/viz-platform/agents/component-check.md b/plugins/viz-platform/agents/component-check.md index 3483e00..2ac1d03 100644 --- a/plugins/viz-platform/agents/component-check.md +++ b/plugins/viz-platform/agents/component-check.md @@ -2,6 +2,8 @@ name: component-check description: DMC component validation specialist model: haiku +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Component Check Agent diff --git a/plugins/viz-platform/agents/design-reviewer.md b/plugins/viz-platform/agents/design-reviewer.md index 58f880b..e552ab0 100644 --- a/plugins/viz-platform/agents/design-reviewer.md +++ b/plugins/viz-platform/agents/design-reviewer.md @@ -2,6 +2,8 @@ name: design-reviewer description: Reviews code for design system compliance using viz-platform MCP tools. Use when validating DMC components, theme tokens, or accessibility standards. model: sonnet +permissionMode: plan +disallowedTools: Write, Edit, MultiEdit --- # Design Reviewer Agent diff --git a/plugins/viz-platform/agents/layout-builder.md b/plugins/viz-platform/agents/layout-builder.md index f5d4760..0066919 100644 --- a/plugins/viz-platform/agents/layout-builder.md +++ b/plugins/viz-platform/agents/layout-builder.md @@ -2,6 +2,7 @@ name: layout-builder description: Practical dashboard layout specialist for creating well-structured layouts with filtering, grid systems, and responsive design. model: sonnet +permissionMode: default --- # Layout Builder Agent diff --git a/plugins/viz-platform/agents/theme-setup.md b/plugins/viz-platform/agents/theme-setup.md index 12be071..6adf94b 100644 --- a/plugins/viz-platform/agents/theme-setup.md +++ b/plugins/viz-platform/agents/theme-setup.md @@ -2,6 +2,7 @@ name: theme-setup description: Design-focused theme setup specialist for creating consistent, brand-aligned themes for Dash Mantine Components applications. model: haiku +permissionMode: acceptEdits --- # Theme Setup Agent