Update the Schema Loading requirement to include all original scenarios
(modified as needed) per the MODIFIED requirement guidelines. The archiver
replaces the entire requirement with the provided content.
Adds two change proposals for implementing Slice 3 of the artifact POC:
1. restructure-schema-directories
- Move schemas from embedded TS objects to self-contained directories
- Each schema becomes a directory with schema.yaml + templates/
- Enables co-located templates for user extensibility
- 2-level resolution: user override → package built-in
2. add-instruction-loader (depends on #1)
- Load templates from schema directories
- Enrich templates with change context (dependencies, next steps)
- Format change status for CLI output
- New instruction-loader capability
These proposals complete Slice 3 from docs/artifact_poc.md.
Add FileSystemUtils.toPosixPath() utility and consolidate scattered
path normalization patterns. This fixes Windows test failures where
fast-glob couldn't match paths containing backslashes.
Root cause: path.join() uses backslashes on Windows, but fast-glob
requires forward slashes for glob patterns on all platforms.
Changes:
- Add toPosixPath() to FileSystemUtils for cross-platform path handling
- Update artifact-graph/state.ts to normalize glob patterns
- Consolidate path normalization in update.ts, validator.ts, and
json-converter.ts to use the new utility
* proposal: add artifact graph core query system
Add OpenSpec change proposal for Slice 1 of the artifact POC - the core
"What's Ready?" query system. This implements:
- ArtifactGraph class for DAG-based dependency modeling
- Filesystem-based state detection (file existence = completion)
- Topological sort for build order calculation
- Ready/blocked artifact queries
This is a parallel module that will coexist with the current system.
* docs: specify Zod for schema validation in artifact graph proposal
- Add decision section for Zod schema validation in design.md
- Update data structures to show Zod schemas with z.infer<> types
- Update tasks to specify Zod usage for type definitions and parsing
* docs: add 2-level schema resolution and built-in schemas
- Add decision for global → built-in schema resolution pattern
- Add resolver.ts for schema lookup logic
- Add built-in schemas directory (spec-driven.yaml, tdd.yaml)
- Add schema resolution tests
- Follows ESLint/Prettier/Git patterns (defaults baked in package)
* experiment: add vertical slice version of artifact graph change
Creates add-artifact-graph-core-v2 with requirements organized as
vertical slices - each requirement file contains its spec, design
decisions, and tasks bundled together for comparison.
* feat(core): add getGlobalDataDir for XDG-compliant data directory
Add getGlobalDataDir() function following XDG Base Directory Specification
for storing user data like schema overrides:
- XDG_DATA_HOME takes precedence on all platforms
- Unix/macOS fallback: ~/.local/share/openspec/
- Windows fallback: %LOCALAPPDATA%/openspec/
* feat(artifact-graph): add core dependency graph module
Implement Slice 1 ("What's Ready?") of the artifact graph system:
- types.ts: Zod schemas for artifact definitions with derived TypeScript types
- schema.ts: YAML parsing with validation for duplicates, invalid refs, cycles
- graph.ts: ArtifactGraph class with Kahn's algorithm for topological sort
- state.ts: Filesystem-based completion detection with glob pattern support
- resolver.ts: Two-level schema resolution (global override → built-in)
- builtin-schemas.ts: spec-driven and tdd workflow definitions
Key design decisions:
- Filesystem as database (stateless, git-friendly)
- Cycle errors show full path (e.g., "A → B → C → A")
- Deterministic ordering via sorted queues
* test(artifact-graph): add comprehensive test suite
52 tests covering all artifact-graph functionality:
- schema.test.ts: Parsing, validation errors, cycle detection
- graph.test.ts: Build order, ready artifacts, blocked queries
- state.test.ts: File existence, glob patterns, missing directories
- resolver.test.ts: Schema resolution with global overrides
* docs(openspec): archive add-artifact-graph-core change
Archive completed change proposal and create artifact-graph spec with
6 requirements covering schema loading, build order, state detection,
ready queries, completion checks, and blocked queries.
* chore: remove experimental artifact-graph-core-v2 folder
Clean up experimental vertical slice proposal that is no longer needed.
* feat(artifact-graph): validate global schema overrides
Global schema overrides are now validated through the same pipeline as
built-in schemas, catching invalid schemas, cyclic dependencies, and
invalid requires references at load time. Added SchemaLoadError for
better error context with file paths.
* test(artifact-graph): add workflow integration tests
Add end-to-end integration tests that exercise the full artifact-graph
pipeline: resolveSchema → ArtifactGraph → detectCompleted → queries.
Tests cover:
- Complete spec-driven and tdd workflow progressions
- Out-of-order file creation handling
- Glob pattern matching with multiple files
- Build order consistency
- Edge cases (empty/missing directories, non-matching files)
* refactor(artifact-graph): adopt zod v4 error message format
Update custom error messages from string format to zod v4 object format
using `{ error: 'message' }` convention.
* fix(test): prevent hanging vitest threads after test runs
- Add teardownTimeout (3s) to vitest config for forced cleanup
- Add global teardown function to vitest.setup.ts
- Call child.unref() to prevent child processes from blocking event loop
- Explicitly destroy stdio streams on process close/error
When creating a new spec file, REMOVED requirements are now ignored
with a warning instead of causing archive to fail. This enables
refactoring scenarios where old fields are removed while documenting
a capability for the first time.
Fixes#403
Add internal documentation for the artifact-based approach to OpenSpec
core. This document outlines design decisions, terminology, and the
philosophy behind treating dependencies as enablers rather than gates.
* fix(cli): respect --no-interactive flag in validate command
The validate command's spinner was starting regardless of the
--no-interactive flag, causing hangs in pre-commit hooks.
Changes:
- Pass noInteractive option to runBulkValidation
- Handle Commander.js --no-* flag syntax (sets interactive=false)
- Only start ora spinner when in interactive mode
- Add CI environment variable check to isInteractive() for industry
standard compliance
* test: add unit tests for interactive utilities and CLI flag
- Export resolveNoInteractive() helper for reuse
- Add InteractiveOptions type export for testing
- Refactor validate.ts to use resolveNoInteractive()
- Add 17 unit tests for isInteractive() and resolveNoInteractive()
- Add CLI integration test for --no-interactive flag
This prevents future regressions where Commander.js --no-* flag
parsing is not properly handled.
* fix(cli): use dynamic import for @inquirer/prompts in config command
The config command (added in #382) reintroduced the pre-commit hook hang
issue that was fixed in #380. The static import of @inquirer/prompts at
module load time causes stdin event listeners to be registered even when
running non-interactive commands, preventing clean process exit when
stdin is piped (as pre-commit does).
Convert the static import to a dynamic import that only loads inquirer
when the `config reset` command is actually used interactively.
Fixes#367
* chore: add ESLint with no-restricted-imports rule for @inquirer
Add ESLint configuration that prevents static imports of @inquirer/*
modules. This prevents future regressions of the pre-commit hook hang
issue fixed in this PR.
The rule shows a helpful error message pointing to issue #367 for context.
init.ts is exempted since it's already dynamically imported from the CLI.
* ci: add ESLint step to lint job
Run `pnpm lint` in CI to enforce the no-restricted-imports rule
that prevents static @inquirer imports.
* feat(cli): add openspec config command for global configuration management
Implements the `openspec config` command with subcommands:
- `path`: Show config file location
- `list [--json]`: Show all current settings
- `get <key>`: Get a specific value (raw output for scripting)
- `set <key> <value> [--string]`: Set a value with auto type coercion
- `unset <key>`: Remove a key (revert to default)
- `reset --all [-y]`: Reset configuration to defaults
- `edit`: Open config in $EDITOR/$VISUAL
Key features:
- Dot notation for nested key access (e.g., featureFlags.someFlag)
- Auto type coercion (true/false → boolean, numbers → number)
- --string flag to force string storage
- Zod schema validation with unknown field passthrough
- Reserved --scope flag for future project-local config
- Windows-compatible editor spawning with proper path quoting
- Shell completion registry integration
* test(config): add additional unit tests for validation and coercion
- Add tests for unknown fields with various types
- Add test to verify error message path for featureFlags
- Add test for number values rejection in featureFlags
- Add config set simulation tests to verify full coerce → set → validate flow
* fix(config): avoid shell parsing in config edit to handle paths with spaces
Use spawn with shell: false and pass configPath as an argument instead
of building a shell command string. This correctly handles spaces in
both the EDITOR path and config file path on all platforms.
* chore(openspec): archive add-config-command and create cli-config spec
Move completed change to archive and apply spec deltas to create
the cli-config specification documenting the config command interface.
* Validate config keys on set
Fixes#367
The CLI was hanging when run as a pre-commit hook because @inquirer/prompts
was statically imported at module load time. Even when prompts were never
called (e.g., `openspec validate --specs --no-interactive`), the import
itself could set up stdin references that prevented clean process exit
when stdin was piped.
Changes:
- Convert all static `@inquirer/prompts` imports to dynamic imports
- Dynamically import `InitCommand` (which uses `@inquirer/core`)
- Update `isInteractive()` to accept options object with both
`noInteractive` and Commander's negated `interactive` property
- Handle empty validation queue with proper exit code
Now when running in non-interactive mode, the inquirer modules are never
loaded, allowing the process to exit cleanly after completion.
Prioritize XDG_CONFIG_HOME on Windows to fix test environment overrides.
Previously, Windows would always use APPDATA regardless of XDG_CONFIG_HOME,
causing tests to fail. Now XDG_CONFIG_HOME is checked first on all platforms
before falling back to platform-specific defaults.
Also update the Windows APPDATA test to explicitly clear XDG_CONFIG_HOME
when testing the fallback behavior.
* feat(core): implement global config directory with XDG support
Add new global-config module following XDG Base Directory Specification with platform-specific fallbacks (Unix: ~/.config/openspec, Windows: %APPDATA%/openspec). Includes config loading with defaults, config saving with directory creation, and full test coverage. Archive add-global-config-dir change.
* docs(spec): add Purpose section to global-config spec
Replace placeholder text with a concise description of what the spec governs, its scope, and high-level objectives.
Create two OpenSpec change proposals:
1. add-global-config-dir: Foundation for user-level configuration following XDG Base Directory Specification with cross-platform support
2. add-config-command: User-facing CLI command for viewing and managing global settings
Both proposals are minimal and focused on providing a clean, extensible base for OpenSpec settings and future feature flags.
- Fix canWriteFile to use fs.access with W_OK flag instead of Unix-style
permission bits (stats.mode & 0o222) which don't work on Windows
- Update test paths to use platform-specific invalid paths that fail on
both Unix and Windows
- Use regex for path separator matching in test assertions
Update the cli-completion spec to match the actual implementation:
- Change `completion zsh` to `completion generate [shell]` command structure
- Update uninstall behavior to reflect confirmation prompt cancels entire operation
- Change "not installed" uninstall exit code from 0 to 1
- Update shell detection error message to match implementation
- Replace Purpose placeholder with actual description
* shell completions for zsh
* after code review changes
* expose only postinstall.js script
* Replace _openspec "$@" with compdef in zsh-generator.ts to prevent execution during load
* Update test/commands/completion.test.ts
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* - Fix dispatcher to use $words instead of $line for subcommand routing
- Add __complete endpoint with tab-separated output for safe parsing
- Replace brittle awk parsing with __complete in completion helpers
- Add uninstall confirmation prompt with --yes flag to skip
- Prefer $ZSH env var for Oh My Zsh detection before dir check
- Add fpath verification guidance for OMZ installations
- Update cli-completion spec to document generate subcommand
* improve shell detection and installation handling
- Return structured result from detectShell() with shell and detected name
- Detect already-installed completions and skip reinstall
- Add update detection with automatic backup of previous version
- Add debug logging to silent catch blocks for diagnostics
- Quote fpath directories to handle paths with spaces
- Verify Oh My Zsh fpath configuration and add to .zshrc if needed
- Show helpful error for detected but unsupported shells
- Update all tests for new detection API
---------
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Co-authored-by: Tabish Bidiwale <30385142+TabishB@users.noreply.github.com>
- Sort supported AI tools alphabetically (A-Z)
- Wrap both "Native Slash Commands" and "AGENTS.md Compatible" sections
in collapsible <details> tags to reduce visual clutter
Remove the `agent: build` field from OpenCode slash command templates
to allow OpenCode to use the current/custom agent instead of requiring
the build agent to be available.
Fixes#334
* Add changeset for new features and improvements
* Update .changeset/new-features-and-improvements.md
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
---------
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Add prominent restart instruction in success message to inform users
they need to restart their IDE/coding tool for slash commands to appear.
Applies to all tools when slash commands are created or refreshed.
Also updates cli-init spec to document the restart instruction requirement.
* support iflow-cli
* docs: add iFlow to supported AI tools in README ([#268](https://github.com/Fission-AI/OpenSpec/pull/268))
Add iFlow to the Native Slash Commands table in the README. iFlow support was implemented but was missing from the documentation.
* add UTs for iflow-cli
* Enhance proposal guidelines in slash-command-templates.ts
- Added instructions to avoid writing code during the proposal stage and focus on creating design documents.
- Emphasized the design phase by including a reminder not to implement code until the apply stage.
- Updated validation steps to clarify the importance of confirming project conventions before proceeding with tasks.
* Update proposal guidelines in slash-command-templates.ts to clarify that no implementation code should be written during the design phase.
* Refine proposal and apply steps in slash-command-templates.ts
- Removed redundant instruction in proposal steps to streamline the process.
- Clarified the initial reading order for apply steps by omitting the project conventions document, focusing on the proposal and design documents instead.
* feat(roocode): Added RooCode tool support and related configurations
Added RooCode tool integration, including:
- Added RooCode configurator class
- Registered RooCode to the tool registry
- Implemented RooCode template files
- Added RooCode slash command support
- Updated README documentation
- Added related test cases
* Removed RooCode related configurations from the project. This includes deleting the RooCode configurator, its template, and associated tests.
* fix(cline): use workflows instead of rules for slash commands
- Update ClineSlashCommandConfigurator to use .clinerules/workflows/ paths
- Update tests to expect correct workflow file locations
- Update README.md to reflect workflows instead of rules
- Fixes Cline integration to match Cline's architecture per their blog post
* Adds spec for fix-cline-workflows-implementation
---------
Co-authored-by: didier <didier.boff@axess.fr>
Co-authored-by: Tabish Bidiwale <30385142+TabishB@users.noreply.github.com>
* feat: add Gemini CLI support with TOML-based slash commands
- Add GeminiSlashCommandConfigurator for .gemini/commands/openspec/
- Register Gemini CLI in AI_TOOLS config and slash command registry
- Generate TOML files with description and prompt fields
- Add comprehensive test coverage for Gemini CLI integration
- Update README to list Gemini CLI under Native Slash Commands
- Remove Gemini CLI from AGENTS.md compatible list (now native)
Implements GitHub issue #248
* [add]
* feat: address PR feedback - remove changeset and add update test
- Remove .changeset/add-gemini-cli-support.md as requested by maintainer
- Add test for Gemini CLI TOML update/refresh path
- Test verifies that existing TOML files are properly updated when running init again
Addresses feedback from TabishB in PR #256
---------
Co-authored-by: Tabish Bidiwale <30385142+TabishB@users.noreply.github.com>
This change adds the $ARGUMENTS property to the OpenCode apply slash command template, allowing users to pass arguments when invoking the apply command. The template now includes instructions for the agent to find and implement the change proposal, with guidance to ask for clarification when ambiguous.
* feat(ai-tools): add Costrict integration support
Add support for Costrict AI tool with slash commands and configuration template. Includes configurator, registry entries, templates, and comprehensive test coverage.
* feat(config): update branding from 'Costrict' to 'CoStrict'
* fix: recreate missing openspec template files in extend mode
The init command now checks for and recreates missing template files
(like openspec/AGENTS.md and openspec/project.md) when running in
extend mode, instead of skipping file generation entirely.
Previously, if a user deleted openspec/AGENTS.md and ran init again,
the file would not be recreated because the openspec/ directory
existed, triggering extend mode which skipped all file generation.
Changes:
- Added ensureTemplateFiles() method to check and recreate missing files
- Modified extend mode to call ensureTemplateFiles() instead of skipping
- Updated message to "Checking for missing files..." for clarity
- Added tests for recreating deleted openspec/AGENTS.md and project.md
* refactor: consolidate duplicate logic in template file generation
Extracted shared logic from generateFiles() and ensureTemplateFiles()
into a new writeTemplateFiles() method with a skipExisting parameter.
This eliminates code duplication while maintaining the same behavior.
* test: improve extend mode test coverage and reduce duplication
- Extracted testFileRecreationInExtendMode helper to reduce code duplication
- Added test to verify existing files are preserved in extend mode
- Ensures skipExisting behavior is properly tested
* fix: prevent false "already configured" detection for tools
Fixes#195
## Problem
Users with existing tool config files (like CLAUDE.md) would see those
tools marked as "already configured" even when running `openspec init`
for the first time. This caused confusion as users thought OpenSpec was
already set up when it wasn't.
Root causes:
1. Tool detection checked only for file existence, not OpenSpec ownership
2. Detection ran even in fresh projects without openspec/ folder
## Solution
Two-part fix:
1. **Conditional detection**: Only check tool configuration when in extend
mode (when openspec/ directory already exists). Fresh initializations
skip the check entirely, treating all tools as unconfigured.
2. **Marker-based validation**: For tools to be considered "configured by
OpenSpec", their files must contain OpenSpec markers (<!-- OPENSPEC:START -->
and <!-- OPENSPEC:END -->). For tools with both config files and slash
commands (like Claude Code), BOTH must have markers.
## Changes
- Modified getExistingToolStates() to accept extendMode parameter
- Rewrote isToolConfigured() to verify OpenSpec markers in files
- Added OPENSPEC_MARKERS to imports
- Added 4 comprehensive test cases covering the new behavior
## Test Coverage
- Fresh init with existing CLAUDE.md → NOT shown as configured ✅
- Fresh init with existing slash commands → NOT shown as configured ✅
- Extend mode with OpenSpec files → shown as configured ✅
- Fresh init with global Codex prompts → NOT shown as configured ✅
All 240 tests pass.
* refactor: optimize tool state detection and improve code clarity
Address code review feedback:
1. **Parallelize tool state checks**: Changed from sequential `for` loop to
`Promise.all()` for checking multiple tools simultaneously. This reduces
I/O latency during extend mode initialization.
2. **Extract marker validation helper**: Created `fileHasMarkers()` helper
function to eliminate code duplication between config file and slash
command checks. Makes the logic clearer and more maintainable.
3. **Clarify slash command policy**: Added explicit comment that "at least
one file with markers is sufficient" (not all required) for slash commands.
This is correct because OpenSpec creates all files together - if any
exists with markers, the tool was configured by OpenSpec.
4. **Simplify fresh init path**: Use `Object.fromEntries()` for cleaner
initialization of all-false states.
Performance improvement: Extend mode now checks tools in parallel instead
of serially, reducing init time especially for projects with many tools.
* fix: use change-id as fallback title instead of "Untitled Change"
Fixes#225 by addressing mismatch between proposal template and title extraction:
- Updated proposal template to include `# Change: [description]` header
- Changed extractTitle fallback from "Untitled Change" to change-id
- Updated all extractTitle call sites to pass changeName parameter
This ensures both new and existing proposals display meaningful titles.
* refactor: make title extraction case-insensitive for "Change:"
Makes the extractTitle regex case-insensitive to handle variations like
"# change:" or "# CHANGE:" in addition to "# Change:".
Tabish Bidiwale