foundry-mcp 0.3.2__tar.gz → 0.7.4__tar.gz

{foundry_mcp-0.3.2 → foundry_mcp-0.7.4}/.gitignore

@@ -51,6 +51,8 @@ settings.json
 
 # Ignore all dotfolders under specs/ (reports, reviews, etc.)
 specs/.*/
+!specs/.plans/
+!specs/.plans/**
 
 # ModelChorus configuration
 .model-chorusrc
@@ -62,3 +64,5 @@ uv.lock
 .venv
 .hypothesis/
 .ruff_cache/
+node_modules/
+.opencode/

foundry_mcp-0.7.4/CHANGELOG.md

@@ -0,0 +1,275 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+### Fixed
+
+- **AI Consultation Config Loading**: Fixed issue where AI consultation features returned `model_used: "none"` because config was loaded from CWD instead of workspace path
+  - `review.py`: Now loads `foundry-mcp.toml` from workspace path for fidelity reviews
+  - `plan.py`: Added `_find_config_file()` helper to walk up directories and find config
+
+### Added
+
+- **Provider Availability Caching**: Cache provider detection results to speed up MCP tool calls
+  - New `[providers] availability_cache_ttl` config option (default: 3600 seconds)
+  - Reduces repeated calls from ~5s to ~0s
+
+### Changed
+
+- **Provider Model Validation Removed**: Model allowlists removed from all CLI providers
+  - Providers no longer pre-register or validate model IDs against hardcoded lists
+  - Any model string is now passed through to the underlying CLI for validation
+  - Eliminates sync issues when providers release new models
+  - Affected providers: `claude`, `gemini`, `codex`, `cursor-agent`
+  - Default models remain as fallbacks: opus, pro, gpt-5.2, composer-1
+
+- **BREAKING: Simplified Spec Templates**: Removed pre-baked spec templates (simple, medium, complex, security)
+  - Only `empty` template is now supported - creates a blank spec with no phases
+  - Use phase templates (`planning`, `implementation`, `testing`, `security`, `documentation`) to add structure
+  - Default template changed from `medium` to `empty`
+  - Mission statement no longer required (was required for medium/complex)
+  - `_requires_rich_task_fields()` now checks explicit `complexity` metadata instead of template
+  - Passing deprecated templates (simple, medium, complex, security) returns validation error
+
+### Migration
+
+```python
+# Old approach (no longer works)
+authoring(action="spec-create", name="my-feature", template="medium", mission="...")
+
+# New approach
+authoring(action="spec-create", name="my-feature")
+authoring(action="phase-template", template_action="apply", template_name="planning", spec_id="...")
+authoring(action="phase-template", template_action="apply", template_name="implementation", spec_id="...")
+```
+
+## [0.7.0] - 2025-12-30
+
+### Added
+
+- **Research Router**: New unified research tool providing multi-model orchestration capabilities
+  - **chat**: Single-model conversation with thread persistence
+    - Thread creation with title and system prompt
+    - Conversation continuation via thread_id
+    - Token budgeting for context management
+    - Thread CRUD operations (list, get, delete)
+  - **consensus**: Multi-model parallel consultation with synthesis
+    - Parallel execution via asyncio.gather with semaphore limiting
+    - Four synthesis strategies: all_responses, synthesize, majority, first_valid
+    - Partial failure handling with min_responses and require_all options
+    - Configurable timeout per provider
+  - **thinkdeep**: Hypothesis-driven systematic investigation
+    - Investigation step execution with state persistence
+    - Hypothesis creation and tracking with evidence accumulation
+    - Confidence level progression (speculation -> confirmed)
+    - Convergence detection based on depth and confidence
+  - **ideate**: Creative brainstorming with idea clustering
+    - Four-phase workflow: divergent, convergent, selection, elaboration
+    - Multi-perspective idea generation
+    - Automatic clustering and scoring
+    - Detailed plan elaboration for selected clusters
+- **ResearchConfig**: New configuration section for research workflows
+  - Configurable storage path, TTL, max messages per thread
+  - Default provider and consensus provider list
+  - ThinkDeep max depth and Ideate perspectives
+- **Research Data Models**: Pydantic models for all workflow states
+  - Enums: WorkflowType, ConfidenceLevel, ConsensusStrategy, ThreadStatus, IdeationPhase
+  - Conversation, ThinkDeep, Ideate, and Consensus state models
+- **File-Based Memory Storage**: Persistent state management for research sessions
+  - FileStorageBackend with CRUD operations
+  - File locking via filelock for thread safety
+  - TTL-based cleanup for expired sessions
+- **Research Test Suite**: 149 tests covering models, memory, and router
+- **Feature Flag**: `research_tools` flag (experimental) gates research tool access
+
+## [0.6.0] - 2025-12-29
+
+### Added
+
+- **Bikelane Intake System**: Fast-capture queue for rapid idea/task capture with automatic triage workflow
+  - **intake-add**: Add items to the intake queue with title, description, priority (p0-p4), tags, source, and requester fields
+    - Idempotency key support for deduplication (checks last 100 items)
+    - Tag normalization to lowercase
+    - Full dry-run support for validation without persistence
+  - **intake-list**: List pending intake items in FIFO order with cursor-based pagination
+    - Configurable page size (1-200, default 50)
+    - Efficient line-hint seeking with fallback to full scan
+    - Returns total_count for queue size visibility
+  - **intake-dismiss**: Mark items as dismissed with optional reason
+    - Atomic file rewrite pattern for data integrity
+    - Supports dry-run mode
+  - JSONL-based storage at `specs/.bikelane/intake.jsonl` with fcntl file locking
+  - Automatic file rotation at 1000 items or 1MB
+  - Thread-safe and cross-process safe with 5-second lock timeout
+  - Security hardening: path traversal prevention, prompt injection sanitization, control character stripping
+  - Feature flag gated: `intake_tools` (experimental, opt-in)
+- **Intake Schema**: JSON Schema for intake-v1 format with comprehensive validation constraints
+- **Intake Documentation**: User guide at `docs/guides/intake.md`
+- **RESOURCE_BUSY Error Code**: New error code for lock contention scenarios
+
+## [0.5.1] - 2025-12-27
+
+### Added
+
+- **Phase Metadata Updates**: New `authoring action=phase-update-metadata` for updating phase-level metadata
+  - Supports updating `estimated_hours`, `description`, and `purpose` fields
+  - Full dry-run support for previewing changes
+  - Tracks previous values for audit purposes
+  - Core function `update_phase_metadata()` in `spec.py` with comprehensive validation
+
+### Fixed
+
+- **Lifecycle Tool Router Compatibility**: Fixed `_handle_move()` and other lifecycle handlers receiving unexpected keyword arguments (`force`, `to_folder`) from the unified router dispatch
+  - All lifecycle handlers now accept full parameter set for router compatibility
+  - Resolves errors like `_handle_move() got an unexpected keyword argument 'force'`
+
+## [0.5.0] - 2025-12-27
+
+### Added
+
+- **Spec Modification Capabilities**: Complete implementation of dynamic spec modification (7 phases, 54 tasks)
+  - **Task Hierarchy Mutations**: `task action=move` for repositioning tasks within/across phases with circular reference prevention
+  - **Dependency Management**: `task action=add-dependency`, `task action=remove-dependency` for blocks/blocked_by/depends relationships
+  - **Task Requirements**: `task action=add-requirement` for adding structured requirements to tasks
+  - **Bulk Operations**: `authoring action=phase-add-bulk` for batch phase creation, `authoring action=phase-template` for applying predefined structures
+  - **Metadata Batch Updates**: `task action=metadata-batch` with AND-based filtering by node_type, phase_id, or pattern regex
+  - **Find-Replace**: `authoring action=spec-find-replace` with regex support and scope filtering for bulk spec modifications
+  - **Spec Rollback**: `authoring action=spec-rollback` for restoring specs from automatic backups
+  - **Spec History & Diff**: `spec action=history` for backup timeline, `spec action=diff` for comparing specs
+  - **Validation Enhancements**: `spec action=completeness-check` with weighted scoring (0-100), `spec action=duplicate-detection` with configurable similarity threshold
+- **Standardized Error Codes**: New `ErrorCode` enum with semantic error codes per 07-error-semantics.md
+- **Contract Tests**: Comprehensive test suite for response-v2 envelope compliance across all phases
+
+### Changed
+
+- Updated capabilities manifest with 15 new actions documented
+- Spec modification spec moved from pending to active (100% complete)
+
+## [0.4.2] - 2025-12-24
+
+### Added
+
+- **Preflight Validation**: `authoring action=spec-create dry_run=true` now generates and validates the full spec, returning `is_valid`, `error_count`, `warning_count`, and detailed diagnostics before actual creation
+- **Schema Introspection**: New `spec action=schema` returns all valid enum values (templates, node_types, statuses, task_categories, verification_types, journal_entry_types, blocker_types, status_folders) for LLM/client discovery
+
+### Changed
+
+- **Spec Field Requirements**: Medium/complex specs now require `metadata.mission` and task metadata for `task_category`, `description`, and `acceptance_criteria`; implementation/refactoring tasks must include `file_path`
+- **Task Metadata Updates**: `task update-metadata` now accepts `acceptance_criteria` and aligns task category validation with the canonical spec categories
+
+## [0.4.1] - 2025-12-24
+
+### Added
+
+- **Batch Metadata Utilities**: New task actions for bulk operations
+  - `task action=metadata-batch`: Apply metadata updates to multiple nodes with AND-based filtering by `node_type`, `phase_id`, or `pattern` regex
+  - `task action=fix-verification-types`: Auto-fix invalid/missing verification types on verify nodes with legacy mapping support
+  - Both actions support `dry_run` mode for previewing changes
+- **Phase-First Authoring**: New `authoring action=phase-add-bulk` for creating multiple phases at once with metadata defaults
+- **Spec Mission Field**: Added `mission` field to spec metadata schema for concise goal statements
+- **Workflow Timeout Override**: AI consultation now supports workflow-specific timeout configuration
+
+### Changed
+
+- **JSON Output Optimization**: CLI and MCP server now emit minified JSON (no indentation) for smaller payloads
+- **Fidelity Review Improvements**: Better path resolution with workspace_root support, graceful handling of non-JSON provider responses
+- **Provider Configuration**: Updated OpenCode model IDs and default model; reordered provider priority
+- **Claude Provider Tests**: Updated to use Haiku model for faster test execution
+
+### Fixed
+
+- Fixed parameter filtering in error_list handler to prevent unexpected argument errors
+- Fixed duplicate file paths in fidelity review implementation artifacts
+- Synced `__init__.py` version with `pyproject.toml`
+
+## [0.4.0] - 2025-12-23
+
+### Changed
+
+- **Verification Types**: Aligned task API and spec validator to use canonical values (`run-tests`, `fidelity`, `manual`)
+  - Task API now accepts `run-tests`, `fidelity`, `manual` (previously `auto`, `manual`, `none`)
+  - Spec validator updated to match canonical schema values
+  - Legacy values automatically mapped: `test` → `run-tests`, `auto` → `run-tests`
+
+### Added
+
+- **Auto-fix for `INVALID_VERIFICATION_TYPE`**: Specs with legacy verification types are now auto-fixable via `validate-fix`
+- **Auto-fix for `INVALID_ROOT_PARENT`**: Specs where spec-root has non-null parent are now auto-fixable
+
+### Removed
+
+- Removed `foundry-mcp-ctl` package and mode-toggling feature - server now always runs with all tools registered
+
+## [0.3.4] - 2025-12-21
+
+_Note: Mode toggling features added in this version were subsequently removed._
+
+## [0.3.3] - 2025-12-17
+
+### Changed
+- **Dashboard**: Refactored pages module to views with cleaner organization
+- **Dashboard**: Improved data stores with better caching and filtering
+- **Observability**: Added action label to tool metrics for router-level granularity
+- **Providers**: Codex CLI now ignores unsupported parameters (warning instead of error)
+
+### Added
+- Dashboard PID file tracking for cross-CLI process management
+- Tool usage dashboard view with action-level breakdown
+- OpenCode Node.js wrapper for subprocess execution
+- Integration tests for provider smoke testing, fidelity review flow, and plan review flow
+
+### Fixed
+- Codex provider environment handling (unsets OPENAI_API_KEY/OPENAI_BASE_URL that interfere with CLI)
+- Minor fixes to Claude and Gemini providers
+
+## [0.3.2] - 2025-12-16
+
+### Added
+- Launcher script (`bin/foundry-mcp`) for configurable Python interpreter selection
+- `FOUNDRY_MCP_PYTHON` environment variable to override the default Python interpreter
+
+### Fixed
+- Removed duplicate `spec_id` and `node_id` fields from task progress response
+
+## [0.3.1] - 2025-12-16
+
+### Removed
+- Removed `code` unified tool (find-class, find-function, callers, callees, trace, impact actions) from MCP surface. Unified manifest reduced from 17 to 16 tools.
+
+## [0.3.0] - 2025-12-15
+
+### Changed
+- Consolidated the MCP tool surface into 17 unified routers (tool + `action`) and aligned CLI/MCP naming.
+- Updated documentation and manifests to reflect the unified router contract.
+
+### Added
+- New completed specs documenting MCP tool consolidation and removal of docquery/rendering/docgen.
+- Unified-manifest budget telemetry (Prometheus metrics, recording rules, alerting rules, and dashboard panels).
+
+### Removed
+- Legacy per-tool MCP modules and legacy CLI command surfaces in favor of unified routers.
+- Docquery/rendering/docgen modules and generated docs previously under `docs/generated/`.
+
+## [0.2.1] - 2025-12-08
+
+### Changed
+- **Dashboard**: Replaced aiohttp+vanilla JS dashboard with Streamlit for better visualizations and interactivity
+- Dashboard dependencies changed from `aiohttp` to `streamlit`, `plotly`, `pandas`
+- Default dashboard port changed from 8080 to 8501 (Streamlit default)
+
+### Added
+- New Streamlit dashboard with 5 pages: Overview, Errors, Metrics, Providers, SDD Workflow
+- Interactive Plotly charts with zoom, pan, and hover tooltips
+- Data export functionality (CSV/JSON download buttons)
+- Cached data access via `@st.cache_data` for performance
+- CLI commands: `dashboard start`, `dashboard stop`, `dashboard status`
+- New SDD Workflow page for spec progress tracking, phase burndown, task status
+- Plan review tool (`plan-review`) for AI-assisted specification review
+
+### Removed
+- Old aiohttp-based dashboard server and static JS/CSS files

{foundry_mcp-0.3.2 → foundry_mcp-0.7.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: foundry-mcp
-Version: 0.3.2
+Version: 0.7.4
 Summary: MCP server for SDD toolkit spec management
 Project-URL: Homepage, https://github.com/tylerburleigh/foundry-mcp
 Project-URL: Repository, https://github.com/tylerburleigh/foundry-mcp
@@ -125,6 +125,11 @@ specs/
 - `plan(action=create|list|review)` supports lightweight planning and review flows.
 - Notifications and sampling channels surface phase completions to MCP clients.
 
+### Batch metadata utilities
+
+- `task(action=metadata-batch)` — Apply metadata updates (e.g., `file_path`, `estimated_hours`) to multiple nodes at once. Supports flexible AND-based filtering by `node_type`, `phase_id`, or `pattern` regex. Includes `dry_run` mode for previewing changes.
+- `task(action=fix-verification-types)` — Auto-fix invalid or missing `verification_type` on verify nodes. Supports legacy mappings (`test` → `run-tests`, `auto` → `run-tests`) and defaults unknown types to `manual`. Includes `dry_run` mode for previewing fixes.
+
 ### Code, docs, and testing intelligence
 
 - Code navigation tools via `code(action=...)` support symbol lookup and call-graph tracing.
@@ -199,6 +204,7 @@ All MCP tools emit the standardized envelope defined in `docs/codebase_standards
 | `FOUNDRY_MCP_API_KEYS` | Comma-separated API keys required for tool access | Disabled |
 | `FOUNDRY_MCP_FEATURE_FLAGS` | Additional feature flags to enable (e.g., `planning_tools`) | Based on spec rollout |
 | `FOUNDRY_MCP_RESPONSE_CONTRACT` | Force response contract version (`v2`) | Auto-negotiated |
+| `FOUNDRY_MODE` | Server mode: `full` (16 tools) or `minimal` (1 wake tool) | `full` |
 | `OPENAI_API_KEY` / `ANTHROPIC_API_KEY` | LLM provider credentials | Not set |
 
 ### TOML configuration

foundry_mcp-0.7.4/PLAN.md

@@ -0,0 +1,96 @@
+# PLAN: Phase-First Spec Authoring Improvements
+
+## Overview
+Deliver three coordinated enhancements that reduce manual MCP calls without relying on Markdown parsing:
+1. **Phase authoring macros** that create entire phases (plus starter tasks/placeholders) in a single command.
+2. **Updated documentation** clarifying the supported schema creation/export workflow.
+3. **Batch metadata utilities** so spec authors can update large sets of tasks/verify nodes at once.
+
+## Goals & Non-Goals
+- **Goals**
+  - Let authors scaffold a whole phase (title, summary, default tasks) with one operation.
+  - Align the sdd-plan documentation with the current set of `spec`/`authoring` commands.
+  - Provide tooling to fix metadata-only validation issues through bulk operations.
+- **Non-Goals**
+  - Markdown/import parsing of plan documents.
+  - Full redesign of task authoring flows; individual task commands still exist for fine tuning.
+  - GUI experiences or IDE plugins.
+
+## Success Metrics
+- A complete phase can be added with <=2 commands (phase macro + optional metadata tweak).
+- Skill docs contain no references to deprecated actions and include copy/pasteable workflows.
+- Batch metadata command can fix all verification/file_path errors in <3 commands even on large specs.
+
+## Work Breakdown
+
+### 1. Phase Authoring Macros (High Priority)
+- **API Design**
+  - Extend authoring plugin with `phase-add-bulk` that accepts a JSON payload describing phase name, summary, and an ordered list of task/verify placeholders.
+  - Allow reuse of predefined phase templates (e.g., “Labeling App Phase 1”) stored with the skill.
+- **Implementation Steps**
+  - Add server-side handler that creates the phase node and loops through provided child definitions, creating tasks/verify nodes atomically.
+  - Support optional metadata defaults (e.g., default owner, spec section) applied to all generated tasks.
+  - Provide dry-run output so authors can inspect the structure before creation.
+- **Skill Integration**
+  - Update sdd-plan commands to surface macros (e.g., `phase add template 1`, `phase add custom` with inline JSON editor).
+  - Include helper snippets for copying a plan section and pasting it into the macro payload.
+- **Testing**
+  - Unit coverage for payload validation (missing titles, duplicated IDs).
+  - Integration test that creates phases 1–3 of the labeling spec using only macro calls.
+- **Example Tool Calls**
+  ```bash
+  mcp__plugin_foundry_foundry-mcp__authoring phase-add-bulk '{"spec_id":"spec-ai-labeling-001","phase":{"phase_id":"phase-3","title":"Integrate Ingestion","summary":"Hook labelers into scoring app"},"children":[{"type":"task","title":"Wire ingestion API","metadata":{"file_path":"src/ingestion/api.py","estimated_hours":3}},{"type":"verify","title":"Regression suite","metadata":{"verification_type":"run-tests","mcp_tool":"mcp__foundry-mcp__test-run"}}]}'
+  mcp__plugin_foundry_foundry-mcp__authoring phase-add-template '{"spec_id":"spec-ai-labeling-001","template_id":"labeling-phase-1","phase_id":"phase-1"}'
+  ```
+
+### 2. Documentation Update for Schema Export Flow (Medium Priority)
+- **Audit** existing `spec`/`authoring` features to ensure docs reflect reality.
+- **Doc Changes**
+  - Remove references to `spec schema-export` in `skills/sdd-plan/SKILL.md` and related references.
+  - Add a “Phase-first authoring” section walking through: create spec from template → run phase macro → fine tune tasks → run metadata batch.
+  - Highlight fallback commands (manual `task add`) for edge cases.
+- **Changelog & Guidance**
+  - Document the rationale for retiring schema-export and link to the new macro workflow in troubleshooting guides.
+- **Example Tool Calls**
+  ```bash
+  mcp__plugin_foundry_foundry-mcp__authoring spec-create '{"name":"spec-ai-labeling-001","template":"medium","category":"implementation"}'
+  mcp__skill_sdd-plan doc-open 'phase-first-authoring'
+  ```
+
+### 3. Batch Metadata Utilities (High Priority)
+- **Command Surface**
+  - Introduce `task metadata-batch` that targets nodes by phase ID, task name regex, or node type.
+  - Support multiple metadata keys per invocation (`file_path`, `verification_type`, owners, labels).
+- **Verification Fixers**
+  - Provide canned operations, e.g., `fix-verification-types phase=verify` to rewrite enums to `test/fidelity` values.
+  - Ensure validator and metadata API share a single source of truth for allowed values.
+- **Safety & UX**
+  - Add dry-run flag and summary output (count of nodes updated, diff preview).
+  - Support rollback on partial failure (transaction or revert copy of previous metadata).
+- **Testing**
+  - Regression tests confirming bulk updates succeed on 50+ tasks and gracefully handle invalid filters.
+- **Example Tool Calls**
+  ```bash
+  mcp__plugin_foundry_foundry-mcp__task metadata-batch '{"spec_id":"spec-ai-labeling-001","filter":{"phase_id":"phase-4","node_type":"task"},"metadata":{"file_path":"src/labeling/phase4.py"}}'
+  mcp__plugin_foundry_foundry-mcp__task metadata-batch '{"spec_id":"spec-ai-labeling-001","filter":{"node_type":"verify"},"metadata":{"verification_type":"fidelity"},"dry_run":true}'
+  mcp__plugin_foundry_foundry-mcp__task metadata-batch '{"spec_id":"spec-ai-labeling-001","filter":{"phase_id":"phase-7","node_type":"verify"},"metadata":{"verification_type":"run-tests"}}'
+  ```
+
+## Dependencies & Risks
+- **Dependencies**
+  - MCP authoring/task routes that can accept batch payloads.
+  - Maintained phase templates stored in the repo/skill.
+- **Risks**
+  - Large payloads might hit RPC limits; mitigate with pagination or per-phase batching.
+  - Batch metadata ops could overwrite intentional differences; dry-run and filter validation are critical.
+  - Documentation must stay in sync as new macros ship; add owner to review doc updates alongside code changes.
+
+## Timeline (Rough)
+1. **Week 1**: Finalize phase macro payload schema; stub batch metadata command; draft doc updates.
+2. **Week 2**: Implement phase macros + tests; build metadata-batch backend with dry-run support.
+3. **Week 3**: Integrate macros into sdd-plan skill, update documentation, run end-to-end scenario tests.
+
+## Validation & Rollout
+- Dogfood on the labeling spec workflow to confirm each phase can be authored in a single macro call plus metadata fix.
+- Update release notes + sdd-plan changelog; provide short Loom/demo if useful.
+- Track telemetry (MCP call counts per spec) to confirm reduction after rollout.

{foundry_mcp-0.3.2 → foundry_mcp-0.7.4}/README.md

@@ -82,6 +82,11 @@ specs/
 - `plan(action=create|list|review)` supports lightweight planning and review flows.
 - Notifications and sampling channels surface phase completions to MCP clients.
 
+### Batch metadata utilities
+
+- `task(action=metadata-batch)` — Apply metadata updates (e.g., `file_path`, `estimated_hours`) to multiple nodes at once. Supports flexible AND-based filtering by `node_type`, `phase_id`, or `pattern` regex. Includes `dry_run` mode for previewing changes.
+- `task(action=fix-verification-types)` — Auto-fix invalid or missing `verification_type` on verify nodes. Supports legacy mappings (`test` → `run-tests`, `auto` → `run-tests`) and defaults unknown types to `manual`. Includes `dry_run` mode for previewing fixes.
+
 ### Code, docs, and testing intelligence
 
 - Code navigation tools via `code(action=...)` support symbol lookup and call-graph tracing.
@@ -156,6 +161,7 @@ All MCP tools emit the standardized envelope defined in `docs/codebase_standards
 | `FOUNDRY_MCP_API_KEYS` | Comma-separated API keys required for tool access | Disabled |
 | `FOUNDRY_MCP_FEATURE_FLAGS` | Additional feature flags to enable (e.g., `planning_tools`) | Based on spec rollout |
 | `FOUNDRY_MCP_RESPONSE_CONTRACT` | Force response contract version (`v2`) | Auto-negotiated |
+| `FOUNDRY_MODE` | Server mode: `full` (16 tools) or `minimal` (1 wake tool) | `full` |
 | `OPENAI_API_KEY` / `ANTHROPIC_API_KEY` | LLM provider credentials | Not set |
 
 ### TOML configuration

{foundry_mcp-0.3.2 → foundry_mcp-0.7.4}/docs/codebase_standards/naming-conventions.md

@@ -20,7 +20,7 @@ Consistent naming shortens discovery time, improves LLM selection accuracy, and
 |--------------|------------------|-------|
 | `environment` | `init`, `verify-env`, `verify-toolchain`, `setup`, `detect` | Workspace + toolchain hygiene |
 | `spec` | `find`, `list`, `validate`, `fix`, `stats`, `validate-fix`, `analyze`, `analyze-deps` | Discovery/validation/analysis |
-| `authoring` | `spec-create`, `spec-template`, `spec-update-frontmatter`, `phase-add`, `phase-remove`, `assumption-add`, `assumption-list`, `revision-add` | Spec mutations |
+| `authoring` | `spec-create`, `spec-template`, `spec-update-frontmatter`, `phase-add`, `phase-add-bulk`, `phase-remove`, `phase-move`, `phase-template`, `phase-update-metadata`, `assumption-add`, `assumption-list`, `revision-add`, `spec-find-replace`, `spec-rollback` | Spec mutations |
 | `task` | `next`, `prepare`, `start`, `complete`, `progress`, `query`, `hierarchy`, `block`, `unblock` | Task execution surface |
 | `lifecycle` | `activate`, `complete`, `archive`, `move`, `state` | Spec folder/state transitions |
 | `journal` | `add`, `list`, `list-unjournaled` | Journal records |