* Stage 6: Complete migration — remove legacy scaffold path (#1924)
Remove the legacy GitHub download and offline scaffold code paths.
All 26 agents now use the integration system exclusively.
Code removal (~1073 lines from __init__.py):
- download_template_from_github(), download_and_extract_template()
- scaffold_from_core_pack(), _locate_release_script()
- install_ai_skills(), _get_skills_dir (restored slim version for presets)
- _has_bundled_skills(), _migrate_legacy_kimi_dotted_skills()
- AGENT_SKILLS_MIGRATIONS, _handle_agent_skills_migration()
- _parse_rate_limit_headers(), _format_rate_limit_error()
- Three-way branch in init() collapsed to integration-only
Config derivation (single source of truth):
- AGENT_CONFIG derived from INTEGRATION_REGISTRY (replaced 180-line dict)
- CommandRegistrar.AGENT_CONFIGS derived from INTEGRATION_REGISTRY (replaced 160-line dict)
- Backward-compat constants kept for presets/extensions: SKILL_DESCRIPTIONS,
NATIVE_SKILLS_AGENTS, DEFAULT_SKILLS_DIR
Release pipeline cleanup:
- Deleted create-release-packages.sh/.ps1 (948 lines of ZIP packaging)
- Deleted create-github-release.sh, generate-release-notes.sh
- Deleted simulate-release.sh, get-next-version.sh, update-version.sh
- Removed .github/workflows/scripts/ directory entirely
- release.yml is now self-contained: check, notes, release all inlined
- Install instructions use uv tool install with version tag
Test cleanup:
- Deleted test_ai_skills.py (tested removed functions)
- Deleted test_core_pack_scaffold.py (tested removed scaffold)
- Cleaned test_agent_config_consistency.py (removed 19 release-script tests)
- Fixed test_branch_numbering.py (removed dead monkeypatches)
- Updated auto-promote tests (verify files created, not tip messages)
1089 tests pass, 0 failures, ruff clean.
* fix: resolve merge conflicts with #2051 (claude as skills)
- Fix circular import: move CommandRegistrar import in claude
integration to inside method bodies (was at module level)
- Lazy-populate AGENT_CONFIGS via _ensure_configs() to avoid
circular import at class definition time
- Set claude registrar_config to .claude/commands (extension/preset
target) since the integration handles .claude/skills in setup()
- Update tests from #2051 to match: registrar_config assertions,
remove --integration tip assertions, remove install_ai_skills mocks
1086 tests pass.
* fix: properly preserve claude skills migration from #2051
Restore ClaudeIntegration.registrar_config to .claude/skills (not
.claude/commands) so extension/preset registrations write to the
correct skills directory.
Update tests that simulate claude setup to use .claude/skills and
check for SKILL.md layout. Some tests still need updating for the
full skills path — 10 remaining failures from the #2051 test
expectations around the extension/preset skill registration flow.
WIP: 1076/1086 pass.
* fix: properly handle SKILL.md paths in extension update rollback and tests
Fix extension update rollback using _compute_output_name() for SKILL.md
agents (converts dots to hyphens in skill directory names). Previously
the backup and cleanup code constructed paths with raw command names
(e.g. speckit.test-ext.hello/SKILL.md) instead of the correct computed
names (speckit-test-ext-hello/SKILL.md).
Test fixes for claude skills migration:
- Update claude tests to use .claude/skills paths and SKILL.md layout
- Use qwen (not claude) for skills-guard tests since claude's agent dir
IS the skills dir — creating it triggers command registration
- Fix test_extension_command_registered_when_extension_present to check
skills path format
1086 tests pass, 0 failures, ruff clean.
* fix: address PR review — lazy init, assertions, deprecated flags
- _ensure_configs(): catch ImportError (not Exception), don't set
_configs_loaded on failure so retries work
- Move _ensure_configs() before unregister loop (not inside it)
- Module-level try/except catches ImportError specifically
- Remove tautology assertion (or True) in test_extensions.py
- Strengthen preset provenance assertion to check source: field
- Mark --offline, --skip-tls, --debug, --github-token as hidden
deprecated no-ops in init()
1086 tests pass.
* fix: remove deleted release scripts from pyproject.toml force-include
Removes force-include entries for create-release-packages.sh/.ps1
which were deleted but still referenced in [tool.hatch.build].
* Use Claude skills for generated commands
* Fix Claude integration and preset skill flows
* Group Claude tests in integration suite
* Align Claude skill frontmatter across generators
* Fix native skill preset cleanup
* Keep legacy AI skills test on legacy path
* Move Claude here-mode test to CLI suite
* Stage 5: Skills, Generic & Option-Driven Integrations (#1924)
Add SkillsIntegration base class and migrate codex, kimi, agy, and
generic to the integration system.
Integrations:
- SkillsIntegration(IntegrationBase) in base.py — creates
speckit-<name>/SKILL.md layout matching release ZIP output byte-for-byte
- CodexIntegration — .agents/skills/, --skills default=True
- KimiIntegration — .kimi/skills/, --skills + --migrate-legacy options,
dotted→hyphenated skill directory migration
- AgyIntegration — .agent/skills/, skills-only (commands deprecated v1.20.5)
- GenericIntegration — user-specified --commands-dir, MarkdownIntegration
- All four have update-context.sh/.ps1 scripts
- All four registered in INTEGRATION_REGISTRY
CLI changes:
- --ai <agent> auto-promotes to integration path for all registered agents
- Interactive agent selection also auto-promotes (bug fix)
- --ai-skills and --ai-commands-dir show deprecation notices on integration path
- Next-steps display shows correct skill invocation syntax for skills integrations
- agy added to CommandRegistrar.AGENT_CONFIGS
Tests:
- test_integration_base_skills.py — reusable mixin with setup, frontmatter,
directory structure, scripts, CLI auto-promote, and complete file inventory
(sh+ps) tests
- Per-agent test files: test_integration_{codex,kimi,agy,generic}.py
- Kimi legacy migration tests, generic --commands-dir validation
- Registry updated with Stage 5 keys
- Removed 9 dead-mock tests, moved 4 integration tests to proper locations
- Fixed all bare project-name tests to use tmp_path
- Fixed 6 pre-existing ANSI escape code test failures in test_extensions.py
and test_presets.py
1524 tests pass, 0 failures.
* fix: remove unused variable flagged by ruff (F841)
* fix: address PR review — integration-type-aware deprecation messages and early generic validation
- --ai-skills deprecation message now distinguishes SkillsIntegration
("skills are the default") from command-based integrations ("has no effect")
- --ai-commands-dir validation for generic runs even when auto-promoted,
giving clear CLI error instead of late ValueError from setup()
- Resolves review comments from #2052
* fix: address PR review round 2
- Remove unused SKILL_DESCRIPTIONS dict from base.py (dead code after
switching to template descriptions for ZIP parity)
- Narrow YAML parse catch from Exception to yaml.YAMLError
- Remove unused shutil import from test_integration_kimi.py
- Remove unused _REGISTRAR_EXEMPT class attr from test_registry.py
- Reword --ai-commands-dir deprecation to be actionable
- Update generic validation error to mention both --ai and --integration
* fix: address PR review round 3
- Clarify parsed_options forwarding is intentional (all options passed,
integrations decide what to use)
- Extract _strip_ansi() helper in test_extensions.py and test_presets.py
- Remove unused pytest import (test_cli.py), unused locals (test_integration_base_skills.py)
- Reword --ai-commands-dir deprecation to be actionable without referencing
the not-yet-implemented --integration-options
* fix: address PR review round 4
- Reorder kimi migration: run super().setup() first so hyphenated
targets exist, then migrate dotted dirs (prevents user content loss)
- Move _strip_ansi() to shared tests/conftest.py, import from there
in test_extensions.py, test_presets.py, test_ai_skills.py
- Remove now-unused re imports from all three test files
* fix: address PR review round 5
- Use write_bytes() for LF-only newlines (no CRLF on Windows)
- Add --integration-options CLI parameter — raw string passed through
to the integration via opts['raw_options']; the integration owns
parsing of its own options
- GenericIntegration.setup() reads --commands-dir from raw_options
when not in parsed_options (supports --integration-options="...")
- Skip early --ai-commands-dir validation when --integration-options
is provided (integration validates in its own setup())
- Remove parse_integration_options from core — integrations parse
their own options
* fix: address PR review round 6
- GenericIntegration is now stateless: removed self._commands_dir
instance state, overrides setup() directly to compute destination
from parsed_options/raw_options on the stack
- commands_dest() raises by design (stateless singleton)
- _quote() in SkillsIntegration now escapes backslashes and double
quotes to produce valid YAML even with special characters
* fix: address PR review round 7
- Support --commands-dir=value form in raw_options parsing (not just
--commands-dir value with space separator)
- Normalize CRLF to LF in write_file_and_record() before encoding
- Persist ai_skills=True in init-options.json when using a
SkillsIntegration, so extensions/presets emit SKILL.md overrides
correctly even without explicit --ai-skills flag
* feat(scripts): add --dry-run flag to create-new-feature scripts
Add a --dry-run / -DryRun flag to both bash and PowerShell
create-new-feature scripts that computes the next branch name,
spec file path, and feature number without creating any branches,
directories, or files. This enables external tools to query the
next available name before running the full specify workflow.
When combined with --json, the output includes a DRY_RUN field.
Without --dry-run, behavior is completely unchanged.
Closes#1931
Assisted-By: 🤖 Claude Code
* fix(scripts): gate specs/ dir creation behind dry-run check
Dry-run was unconditionally creating the root specs/ directory via
mkdir -p / New-Item before the dry-run guard. This violated the
documented contract of zero side effects. Also adds returncode
assertion on git branch --list in tests and adds PowerShell dry-run
test coverage (skipped when pwsh unavailable).
Addresses review comments on #1998.
Assisted-By: 🤖 Claude Code
* fix: address PR review feedback
- Gate `mkdir -p $SPECS_DIR` behind DRY_RUN check (bash + PowerShell)
so dry-run creates zero directories
- Add returncode assertion on `git branch --list` in test
- Strengthen spec dir test to verify root `specs/` is not created
- Add PowerShell dry-run test class (5 tests, skipped without pwsh)
- Fix run_ps_script to use temp repo copy instead of project root
Assisted-By: 🤖 Claude Code
* fix: use git ls-remote for remote-aware dry-run numbering
Dry-run now queries remote branches via `git ls-remote --heads`
(read-only, no fetch) to account for remote-only branches when
computing the next sequential number. This prevents dry-run from
returning a number that already exists on a remote.
Added test verifying dry-run sees remote-only higher-numbered
branches and adjusts numbering accordingly.
Assisted-By: 🤖 Claude Code
* fix(scripts): deduplicate number extraction and branch scanning logic
Extract shared _extract_highest_number helper (bash) and
Get-HighestNumberFromNames (PowerShell) to eliminate duplicated
number extraction patterns between local branch and remote ref
scanning.
Add SkipFetch/skip_fetch parameter to check_existing_branches /
Get-NextBranchNumber so dry-run reuses the same function instead
of inlining duplicate max-of-branches-and-specs logic.
Assisted-By: 🤖 Claude Code
* fix(tests): use isolated paths for remote branch test
Move remote.git and second_clone directories under git_repo
instead of git_repo.parent to prevent path collisions with
parallel test workers.
Assisted-By: 🤖 Claude Code
* fix: address PR review feedback
- Set GIT_TERMINAL_PROMPT=0 for git ls-remote calls to prevent
credential prompts from blocking dry-run in automation scenarios
- Add returncode assertion to test_dry_run_with_timestamp git
branch --list check
Assisted-By: 🤖 Claude Code
* fix: support feature branch numbers with 4+ digits in common.sh and common.ps1
The sequential feature number pattern was hardcoded to exactly 3 digits
(`{3}`), causing branches like `1234-feature-name` to be rejected.
Changed to `{3,}` (3 or more digits) to support growing projects.
Also added a guard to exclude malformed timestamp patterns from being
accepted as sequential prefixes.
Closes#344
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: narrow timestamp guard and use [long] to prevent overflow
- Change [int] to [long] in PowerShell Get-CurrentBranch to avoid
overflow for large feature numbers (>2,147,483,647)
- Narrow malformed-timestamp exclusion from ^[0-9]+-[0-9]{6}- to
^[0-9]{7}-[0-9]{6}- so valid sequential branches like
004-123456-fix-bug are not rejected
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* test: add regression tests for 4+ digit feature branch support
Cover check_feature_branch and find_feature_dir_by_prefix with 4-digit
sequential prefixes, as requested in PR review #2040.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: reject timestamp-like branches without trailing slug
Branches like "20260319-143022" (no "-<name>" suffix) were incorrectly
accepted as sequential prefixes. Add explicit rejection for 7-or-8
digit date + 6-digit time patterns with no trailing slug, in both
common.sh and common.ps1.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Add community content disclaimers
Add notes clarifying that community extensions, presets, walkthroughs,
and community friends are independently created and maintained by their
respective authors and are not reviewed, nor endorsed, nor supported
by GitHub.
Disclaimers added to:
- README.md: Community Extensions, Community Presets, Community
Walkthroughs, and Community Friends sections
- extensions/README.md: Community Reference Catalog and Available
Community Extensions sections
- presets/README.md: Catalog Management section
* Refine community disclaimers per PR review feedback
- Clarify that GitHub/maintainers may review catalog PRs for formatting
and policy compliance, but do not review, audit, endorse, or support
the extension/preset code itself (avoids contradiction with submission
process that mentions PR reviews)
- Add missing 'use at your own discretion' guidance to Community
Walkthroughs and Community Friends sections for consistency
* docs: remove dead Cognitive Squad and Understanding extension links
Both repos (Testimonial/cognitive-squad and Testimonial/understanding)
have been deleted by their author. No forks or relocations exist.
* chore: remove dead extensions from community catalog
Remove cognitive-squad and understanding entries whose repos
have been deleted by their author.
* Stage 3: Standard markdown integrations — 19 agents migrated to plugin architecture
Migrate all standard markdown integrations to self-contained subpackages
under integrations/. Each subclasses MarkdownIntegration with config-only
overrides (~10 lines per __init__.py).
Integrations migrated (19):
claude, qwen, opencode, junie, kilocode, auggie, roo, codebuddy,
qodercli, amp, shai, bob, trae, pi, iflow, kiro-cli, windsurf,
vibe, cursor-agent
Changes:
- Create integrations/<key>/ subpackage with __init__.py and scripts/
(update-context.sh, update-context.ps1) for each integration
- Register all 19 in INTEGRATION_REGISTRY (20 total with copilot)
- MarkdownIntegration.setup() processes templates (replaces {SCRIPT},
{ARGS}, __AGENT__; strips frontmatter blocks; rewrites paths)
- Extract install_scripts() to IntegrationBase; refactor copilot to use it
- Generalize --ai auto-promote from copilot-only to registry-driven:
any integration registered in INTEGRATION_REGISTRY auto-promotes.
Unregistered agents (gemini, tabnine, codex, kimi, agy, generic)
continue through the legacy --ai path unchanged.
- Fix cursor/cursor-agent key mismatch in CommandRegistrar.AGENT_CONFIGS
- Add missing vibe entry to CommandRegistrar.AGENT_CONFIGS
- Update kiro alias test to reflect auto-promote behavior
Testing:
- Per-agent test files (test_integration_<agent>.py) with shared mixin
- 1316 tests passing, 0 failures
- Complete file inventory tests for both sh and ps variants
- Byte-for-byte validated against v0.4.3 release packages (684 files)
* Address PR review: fix repo root detection and no-op test
- Fix repo root fallback in all 20 update-context.sh scripts: walk up
from script location to find .specify/ instead of falling back to pwd
- Fix repo root fallback in all 20 update-context.ps1 scripts: walk up
from script location to find .specify/ instead of falling back to $PWD
- Add assertions to test_setup_writes_to_correct_directory: verify
expected_dir exists and all command files reside under it
* Fix REPO_ROOT priority: prefer .specify walk-up over git root
In monorepos the git toplevel may differ from the project root that
contains .specify/. The previous fix still preferred git rev-parse
over the walk-up result.
Bash scripts (20): prefer the discovered _root when it contains
.specify/; only accept git root if it also contains .specify/.
PowerShell scripts (20): validate git root contains .specify/ before
using it; fall back to walking up from script directory otherwise.
* Guard git call with try/catch in PowerShell scripts
With $ErrorActionPreference = 'Stop', an unguarded git rev-parse
throws a terminating CommandNotFoundException when git is not
installed, preventing the .specify walk-up fallback from running.
Wrap the git call in try/catch across all 20 update-context.ps1
scripts so the fallback works reliably without git.
* Rename hyphenated package dirs to valid Python identifiers
Rename kiro-cli → kiro_cli and cursor-agent → cursor_agent so the
packages can be imported with normal Python syntax instead of
importlib. The user-facing integration key (IntegrationBase.key)
stays hyphenated to match the actual CLI tool / binary name.
Also reorganize _register_builtins(): imports and registrations
are now grouped alphabetically with clear section comments.
* Reuse CommandRegistrar path rewriting in process_template()
Replace the duplicated regex-based path rewriting in
MarkdownIntegration.process_template() with a call to the shared
CommandRegistrar._rewrite_project_relative_paths() implementation.
This ensures extension-local paths are preserved and boundary rules
stay consistent across the codebase.
* Promote _rewrite_project_relative_paths to public API
Rename CommandRegistrar._rewrite_project_relative_paths() to
rewrite_project_relative_paths() (drop leading underscore) so
integrations can call it without reaching into a private method
across subsystem boundaries.
Addresses PR review feedback:
https://github.com/github/spec-kit/pull/2038#discussion_r3022105627
* Broaden TestRegistrarKeyAlignment to cover all integration keys
Parametrize across ALL_INTEGRATION_KEYS instead of only checking
cursor-agent and vibe. Keeps a separate negative test for the
stale 'cursor' shorthand.
Addresses PR review feedback:
https://github.com/github/spec-kit/pull/2038#discussion_r3022269032
* chore: bump version to 0.4.4
* chore: begin 0.4.5.dev0 development
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* feat: Stage 2a — CopilotIntegration with shared template primitives
- base.py: added granular primitives (shared_commands_dir,
shared_templates_dir, list_command_templates, command_filename,
commands_dest, copy_command_to_directory, record_file_in_manifest,
write_file_and_record, process_template)
- CopilotIntegration: uses primitives to produce .agent.md commands,
companion .prompt.md files, and .vscode/settings.json
- Verified byte-for-byte parity with old release script output
- Copilot auto-registered in INTEGRATION_REGISTRY
- 70 tests (22 new: base primitives + copilot integration)
Part of #1924
* feat: Stage 2b — --integration flag, routing, agent.json, shared infra
- Added --integration flag to init() (mutually exclusive with --ai)
- --ai copilot auto-promotes to integration path with migration nudge
- Integration setup writes .specify/agent.json with integration key
- _install_shared_infra() copies scripts and templates to .specify/
- init-options.json records 'integration' key when used
- 4 new CLI tests: mutual exclusivity, unknown rejection, copilot
end-to-end, auto-promote (74 total integration tests)
Part of #1924
* feat: Stage 2 completion — integration scripts, integration.json, shared manifest
- Added copilot/scripts/update-context.sh and .ps1 (thin wrappers
that delegate to the shared update-agent-context script)
- CopilotIntegration.setup() installs integration scripts to
.specify/integrations/copilot/scripts/
- Renamed agent.json → integration.json with script paths
- _install_shared_infra() now tracks files in
integration-shared.manifest.json
- Updated tests: scripts installed, integration.json has script paths,
shared manifest recorded (74 tests)
Part of #1924
* refactor: rename shared manifest to speckit.manifest.json
Cleaner naming — the shared infrastructure (scripts, templates)
belongs to spec-kit itself, not to any specific integration.
* fix: copilot update-context scripts reflect target architecture
Scripts now source shared functions (via SPECKIT_SOURCE_ONLY=1) and
call update_agent_file directly with .github/copilot-instructions.md,
rather than delegating back to the shared case statement.
* fix: simplify copilot scripts — dispatcher sources common functions
Integration scripts now contain only copilot-specific logic (target
path + agent name). The dispatcher is responsible for sourcing shared
functions before calling the integration script.
* fix: copilot update-context scripts are self-contained implementations
These scripts ARE the implementation — the dispatcher calls them.
They source common.sh + update-agent-context functions, gather
feature/plan data, then call update_agent_file with the copilot
target path (.github/copilot-instructions.md).
* docs: add Stage 7 activation note to copilot update-context scripts
* test: add complete file inventory test for copilot integration
Validates every single file (37 total) produced by
specify init --integration copilot --script sh --no-git.
* test: add PowerShell file inventory test for copilot integration
Validates all 37 files produced by --script ps variant, including
.specify/scripts/powershell/ instead of bash.
* refactor: split test_integrations.py into tests/integrations/ directory
- test_base.py: IntegrationOption, IntegrationBase, MarkdownIntegration, primitives
- test_manifest.py: IntegrationManifest, path traversal, persistence, validation
- test_registry.py: INTEGRATION_REGISTRY
- test_copilot.py: CopilotIntegration unit tests
- test_cli.py: --integration flag, auto-promote, file inventories (sh + ps)
- conftest.py: shared StubIntegration helper
76 integration tests + 48 consistency tests = 124 total, all passing.
* refactor: move file inventory tests from test_cli to test_copilot
File inventories are copilot-specific. test_cli.py now only tests
CLI flag mechanics (mutual exclusivity, unknown rejection, auto-promote).
* fix: skip JSONC merge to preserve user settings, fix docstring
- _merge_vscode_settings() now returns early (skips merge) when
existing settings.json can't be parsed (e.g. JSONC with comments),
instead of overwriting with empty settings
- Updated _install_shared_infra() docstring to match implementation
(scripts + templates, speckit.manifest.json)
* fix: warn user when JSONC settings merge is skipped
* fix: show template content when JSONC merge is skipped
User now sees the exact settings they should add manually.
* fix: document process_template requirement, merge scripts without rmtree
- base.py setup() docstring now explicitly states raw copy behavior
and directs to CopilotIntegration for process_template example
- _install_shared_infra() uses merge/overwrite instead of rmtree to
preserve user-added files under .specify/scripts/
* fix: don't overwrite pre-existing shared scripts or templates
Only write files that don't already exist — preserves any user
modifications to shared scripts (common.sh etc.) and templates.
* fix: warn user about skipped pre-existing shared files
Lists all shared scripts and templates that were not copied because
they already existed in the project.
* test: add test for shared infra skip behavior on pre-existing files
Verifies that _install_shared_infra() preserves user-modified scripts
and templates while still installing missing ones.
* fix: address review — containment check, deterministic prompts, manifest accuracy
- CopilotIntegration.setup() adds dest containment check (relative_to)
- Companion prompts generated from templates list, not directory glob
- _install_shared_infra() only records files actually copied (not pre-existing)
- VS Code settings tests made unconditional (assert template exists)
- Inventory tests use .as_posix() for cross-platform paths
* fix: correct PS1 function names, document SPECKIT_SOURCE_ONLY prerequisite
- Fixed Get-FeaturePaths → Get-FeaturePathsEnv, Read-PlanData → Parse-PlanData
- Documented that shared scripts must guard Main with SPECKIT_SOURCE_ONLY
before these integration scripts can be activated (Stage 7)
* fix: add dict type check for settings merge, simplify PS1 to subprocess
- _merge_vscode_settings() skips merge with warning if parsed JSON
is not a dict (array, null, etc.)
- PS1 update-context.ps1 uses & invocation instead of dot-sourcing
since the shared script runs Main unconditionally
* fix: skip-write on no-op merge, bash subprocess, dynamic integration list
- _merge_vscode_settings() only writes when keys were actually added
- update-context.sh uses exec subprocess like PS1 version
- Unknown integration error lists available integrations dynamically
* fix: align path rewriting with release script, add .specify/.specify/ fix
Path rewrite regex matches the release script's rewrite_paths()
exactly (verified byte-identical output). Added .specify/.specify/
double-prefix fix for additional safety.
* docs: sync AGENTS.md with AGENT_CONFIG for missing agents
Add Antigravity (agy) and Mistral Vibe (vibe) to the supported agents
table — both exist in AGENT_CONFIG but were missing from documentation.
Fix Agent Categories section:
- Move Cursor from CLI-Based to IDE-Based (requires_cli is False)
- Add missing CLI agents: Codex, Auggie, iFlow
- Add missing IDE agents: Kilo Code, Roo Code, Trae, Antigravity
Update Command File Formats and Directory Conventions sections to
include all agents that were previously undocumented.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: address Copilot review feedback on AGENTS.md
- Fix Cursor table entry: CLI Tool → N/A (IDE-based), matches requires_cli=False in AGENT_CONFIG
- Fix Antigravity directory: .agent/commands/ → .agent/skills/ (skills-based per AGENT_SKILLS_MIGRATIONS)
- Add opencode singular command exception to Directory Conventions (.opencode/command/)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: add --ai key hint for Cursor in AGENTS.md
Cursor's AGENT_CONFIG key is cursor-agent but the CLI Tool column
shows N/A (IDE-based). Adding the --ai flag reference in the
Description column so readers know the correct key to use with
specify init --ai cursor-agent.
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: add --ai key hint for Antigravity in AGENTS.md
Antigravity's AGENT_CONFIG key is 'agy' and requires --ai-skills flag,
but the table only showed N/A (IDE-based). Adding the --ai flag reference
so readers know to use: specify init --ai agy --ai-skills
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: clarify Antigravity directory convention for both modes
AGENT_CONFIG generates .agent/commands/ by default, but --ai-skills
uses .agent/skills/. Document both paths in Directory Conventions.
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: add Tabnine nested path exception to Directory Conventions
Tabnine uses .tabnine/agent/commands/ which has an extra path segment
compared to the usual .<agent-name>/commands/ convention.
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: add Copilot to Markdown list, fix Antigravity dir convention
- Add GitHub Copilot to the Markdown format "Used by" list (it uses
markdown with .agent.md extension and chat mode frontmatter)
- Clarify Antigravity uses .agent/skills/ (requires --ai-skills);
.agent/commands/ is deprecated/legacy
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: add missing IDE agents to Directory Conventions
Add Roo Code and IBM Bob to the IDE agents list in Directory
Conventions so all IDE-based agents are documented.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: document Codex --ai-skills requirement in all sections
Codex CLI requires --ai-skills when explicitly selected via
specify init --ai codex (exits with migration error otherwise).
Updated table, CLI-Based Agents list, and Directory Conventions.
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: fix Antigravity dir to match AGENT_CONFIG, add Amp shared folder
- Antigravity table row: .agent/skills/ → .agent/commands/ (matches
AGENT_CONFIG folder + commands_subdir; skills mode via --ai-skills)
- Add shared .agents/ folder exception for Amp and Codex
- Move Codex from Skills-based to Shared folder section (it shares
.agents/ with Amp)
Addresses Copilot review feedback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* docs: clarify Antigravity skills path is required, not optional
Reword to make clear .agent/skills/ is the effective path and
.agent/commands/ is deprecated, since CLI enforces --ai-skills.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat: Stage 1 — integration foundation (base classes, manifest, registry)
Add the integrations package with:
- IntegrationBase ABC and MarkdownIntegration base class
- IntegrationOption dataclass for per-integration CLI options
- IntegrationManifest with SHA-256 hash-tracked install/uninstall
- INTEGRATION_REGISTRY (empty, populated in later stages)
- 34 tests at 98% coverage
Purely additive — no existing code modified.
Part of #1924
* fix: normalize manifest keys to POSIX, type manifest parameter
- Store manifest file keys using as_posix() after resolving relative
to project root, ensuring cross-platform portable manifests
- Type the manifest parameter as IntegrationManifest (via TYPE_CHECKING
import) instead of Any in IntegrationBase methods
* fix: symlink safety in uninstall/setup, handle invalid JSON in load
- uninstall() now uses non-resolved path for deletion so symlinks
themselves are removed, not their targets; resolve only for
containment validation
- setup() keeps unresolved dst_file for copy; resolves separately
for project-root validation
- load() catches json.JSONDecodeError and re-raises as ValueError
with the manifest path for clearer diagnostics
- Added test for invalid JSON manifest loading
* fix: lexical symlink containment, assert project_root consistency
- uninstall() now uses os.path.normpath for lexical containment check
instead of resolve(), so in-project symlinks pointing outside are
still properly removed
- setup() asserts manifest.project_root matches the passed project_root
to prevent path mismatches between file operations and manifest
recording
* fix: handle non-files in check_modified/uninstall, validate manifest key
- check_modified() treats non-regular-files (dirs, symlinks) as modified
instead of crashing with IsADirectoryError
- uninstall() skips directories (adds to skipped list), only unlinks
files and symlinks
- load() validates stored integration key matches the requested key
* fix: safe symlink handling in uninstall
- Broken symlinks now removable (lexists check via is_symlink fallback)
- Symlinks never hashed (avoids following to external targets)
- Symlinks only removed with force=True, otherwise skipped
* fix: robust unlink, fail-fast config validation, symlink tests
- uninstall() wraps path.unlink() in try/except OSError to avoid
partial cleanup on race conditions or permission errors
- setup() raises ValueError on missing config or folder instead of
silently returning empty
- Added 3 tests: symlink in check_modified, symlink skip/force in
uninstall (47 total)
* fix: check_modified uses lexical containment, explicit is_symlink check
- check_modified() no longer calls _validate_rel_path (which resolves
symlinks); uses lexical checks (is_absolute, '..' in parts) instead
- is_symlink() checked before is_file() so symlinks to files are still
treated as modified
- Fixed templates_dir() docstring to match actual behavior
---------
Co-authored-by: Manfred Riem <15701806+mnriem@users.noreply.github.com>
* chore: use PEP 440 .dev0 versions on main after releases
- Release-trigger workflow now adds a dev bump commit (X.Y.(Z+1).dev0)
on the release branch after tagging, so main gets the dev version
when the PR merges. The tag still points at the release commit.
- Set current pyproject.toml to 0.4.4.dev0.
- Replace broken release workflow badge with shields.io release badge.
* Update .github/workflows/release-trigger.yml
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* docs: correct specify extension add syntax to require extension name
The specify extension add command requires the extension name as a positional argument. Many documentation files incorrectly demonstrated using the --from flag without specifying the extension name first.
* feat: add superb extension to community catalog
Orchestrates obra/superpowers skills within the spec-kit SDD workflow.
* fix: link superb extension docs
* feat(scripts): add --allow-existing-branch flag to create-new-feature
Add an --allow-existing-branch / -AllowExistingBranch flag to both
bash and PowerShell create-new-feature scripts. When the target branch
already exists, the script switches to it instead of failing. The spec
directory and template are still created if missing, but existing
spec.md files are not overwritten (prevents data loss on re-runs).
The flag is opt-in, so existing behavior is completely unchanged
without it. This enables worktree-based workflows and CI/CD pipelines
that create branches externally before running speckit.specify.
Relates to #1931. Also addresses #1680, #841, #1921.
Assisted-By: 🤖 Claude Code
* fix: address PR review feedback for allow-existing-branch
- Make checkout failure fatal instead of suppressing with || true (bash)
- Check $LASTEXITCODE after git checkout in PowerShell
- Use Test-Path -PathType Leaf for spec file existence check (PS)
- Add PowerShell static assertion test for -AllowExistingBranch flag
Assisted-By: 🤖 Claude Code
* Fix Claude Code CLI detection for npm-local installs
`specify check` reports "Claude Code CLI (not found)" for users who
installed Claude Code via npm-local (the default installer path, common
with nvm). The binary lives at ~/.claude/local/node_modules/.bin/claude
which was not checked. Add CLAUDE_NPM_LOCAL_PATH as a second well-known
location alongside the existing migrate-installer path.
Fixes https://github.com/github/spec-kit/issues/550
* Address Copilot review feedback
- Remove unused pytest import from test_check_tool.py
- Use tmp_path instead of hardcoded /nonexistent/claude for hermetic tests
- Simplify redundant exists() + is_file() to just is_file()
AI-assisted: Changes applied with Claude Code.
* Update tests/test_check_tool.py
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Update tests/test_check_tool.py
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Rename the Normalize-List parameter in create-release-packages.ps1 to avoid conflicting with PowerShell's automatic $input variable. This fixes Windows offline scaffolding when -Agents and -Scripts are passed.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
* feat: add MAQA extension suite to community catalog and README
Adds 7 extensions forming the MAQA (Multi-Agent & Quality Assurance)
suite to catalog.community.json in correct alphabetical order (after
'learn', before 'onboard') and to the README community extensions table:
- maqa — coordinator/feature/QA workflow, board auto-detection
- maqa-azure-devops — Azure DevOps Boards integration
- maqa-ci — CI/CD gate (GitHub Actions/CircleCI/GitLab/Bitbucket)
- maqa-github-projects — GitHub Projects v2 integration
- maqa-jira — Jira integration
- maqa-linear — Linear integration
- maqa-trello — Trello integration
All entries placed alphabetically. maqa v0.1.3 bumped to reflect
multi-board auto-detection added in this release.
* fix: set catalog updated_at to match latest entry timestamp
Top-level updated_at was 00:00:00Z while plan-review-gate entries
had 08:22:30Z, making metadata inconsistent for freshness consumers.
Updated to 2026-03-27T08:22:30Z (>= all entry timestamps).
- Extension ID: plan-review-gate
- Version: 1.0.0
- Author: luno
- Catalog entries sorted alphabetically by ID
- README table row inserted alphabetically by name
Co-authored-by: Ed Harrod <your-real-email@luno.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
The `?.` (null-conditional member access) operator requires PowerShell 7.1+,
but Windows ships with PowerShell 5.1 by default. When AI agents invoke .ps1
scripts on Windows, they typically use the system-associated handler (5.1),
causing a ParseException: Unexpected token '?.Path'.
Replace the single `?.` usage with a 5.1-compatible two-step pattern that
preserves the same null-safety behavior.
Fixes#1972
* chore: bump version to 0.4.2
* chore: clean up CHANGELOG and fix release workflow
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* docs: add manual testing guide for slash command validation
Adds a top-level TESTING.md that describes the manual test process PR
submitters must follow when their changes affect slash commands.
Includes:
- Process overview (identify affected commands, setup, run, report)
- Local setup instructions using editable install
- Reporting template for PR submissions
- Agent prompt that analyzes changed files and determines which
commands need testing, including transitive script dependencies
and extension hook mappings
* Update TESTING.md
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Update TESTING.md
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Add AIDE, Extensify, and Presetify to community extensions
Add three extensions from the mnriem/spec-kit-extensions repository:
- AI-Driven Engineering (AIDE): structured 7-step workflow for building
new projects from scratch with AI assistants
- Extensify: create and validate extensions and extension catalogs
- Presetify: create and validate presets and preset catalogs
Updates both the README community extensions table and
catalog.community.json with entries in alphabetical order.
* fix(tests): isolate preset search test from community catalog growth
Mock get_active_catalogs to return only the default catalog entry so
the test uses only its own cached data and won't break as the
community preset catalog grows.
- Add 🎨 Community Presets section between Community Extensions and Community Walkthroughs
- Add ToC entry for the new section
- Populate presets/catalog.community.json with pirate and aide-in-place presets
- Entries alphabetized: catalog by id, README table by name
- Add 🧩 Community Extensions section to README.md before Community Walkthroughs
- Add table of contents entry for the new section
- Replace extensions/README.md table with a link back to the main README
- Update EXTENSION-PUBLISHING-GUIDE.md references to point to README.md
- Update EXTENSION-DEVELOPMENT-GUIDE.md references to point to README.md
* docs(readme): consolidate Community Friends sections and fix ToC anchors
- Merge duplicate 🤝 Community Friends section (table format near bottom) into
the existing 🛠️ Community Friends section (bullet list)
- Add cc-sdd entry alongside Spec Kit Assistant
- Update intro text to 'Community projects that extend, visualize, or build on Spec Kit'
- Fix ToC anchors for Video Overview and Community Friends (remove variation selector from fragment)
* docs(readme): remove stale ToC entry for deleted Community Friends section
* fix(commands): rename NFR references to success criteria in analyze and clarify
* fix(analyze): align Success Criteria description and inventory keys with spec template
- Reword "non-functional targets" to "measurable outcomes" to match the spec template's broader scope (performance, user success, business impact)
- Use explicit FR-/SC- identifiers as primary stable keys in the requirements inventory instead of derived slugs alone
* Add Community Friends section with cc-sdd
Adds a new "Community Friends" section to the README for projects that
extend or build on Spec Kit. Starts with cc-sdd, a Claude Code plugin
that layers composable traits (quality gates, worktree isolation, agent
teams) on top of Spec Kit's core workflow.
Suggested by @mnriem in discussion #1889.
Assisted-By: 🤖 Claude Code
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Update cc-sdd repo URL after rename
Assisted-By: 🤖 Claude Code
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Mention Superpowers explicitly in cc-sdd description
Assisted-By: 🤖 Claude Code
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(scripts): prioritize .specify over git for repo root detection
When spec-kit is initialized in a subdirectory that doesn't have its
own .git, but a parent directory does, spec-kit was incorrectly using
the parent's git repository root. This caused specs to be created in
the wrong location.
The fix changes repo root detection to prioritize .specify directory
over git rev-parse, ensuring spec-kit respects its own initialization
boundary rather than inheriting a parent git repo.
Fixes#1932
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: address code review feedback
- Normalize paths in find_specify_root to prevent infinite loop with relative paths
- Use -PathType Container in PowerShell to only match .specify directories
- Improve has_git/Test-HasGit to check git command availability and validate work tree
- Handle git worktrees/submodules where .git can be a file
- Remove dead fallback code in create-new-feature scripts
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: check .specify before termination in find_specify_root
Fixes edge case where project root is at filesystem root (common in
containers). The loop now checks for .specify before checking the
termination condition.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: scope git operations to spec-kit root & remove unused helpers
- get_current_branch now uses has_git check and runs git with -C to
prevent using parent git repo branch names in .specify-only projects
- Same fix applied to PowerShell Get-CurrentBranch
- Removed unused find_repo_root() from create-new-feature.sh
- Removed unused Find-RepositoryRoot from create-new-feature.ps1
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: use cd -- to handle paths starting with dash
Prevents cd from interpreting directory names like -P or -L as options.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: check git command exists before calling get_repo_root in has_git
Avoids unnecessary work when git isn't installed since get_repo_root
may internally call git rev-parse.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(powershell): use LiteralPath and check git before Get-RepoRoot
- Use -LiteralPath in Find-SpecifyRoot to handle paths with wildcard
characters ([, ], *, ?)
- Check Get-Command git before calling Get-RepoRoot in Test-HasGit to
avoid unnecessary work when git isn't installed
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(powershell): use LiteralPath for .git check in Test-HasGit
Prevents Test-Path from treating wildcard characters in paths as globs.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(powershell): use LiteralPath in Get-RepoRoot fallback
Prevents Resolve-Path from treating wildcard characters as patterns.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: iamaeroplane <michal.bachorik@gmail.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
