fix: allow Claude to chain skills for hook execution by mnriem · Pull Request #2227 · github/spec-kit

mnriem · 2026-04-15T13:53:15Z

Summary

Fixes #2178 — Claude Code fails to execute speckit commands because:

disable-model-invocation: true prevents Claude from invoking extension skills (e.g., speckit-git-feature) when chaining from workflow skills
Hook sections reference dot-notation command names (speckit.git.commit) but Claude skills use hyphens (speckit-git-commit)
Unicode ✓ in PowerShell auto-commit.ps1 causes parser errors on Windows

Changes

Core fix: `disable-model-invocation` → `false`

Removed Claude-specific frontmatter from the shared build_skill_frontmatter() in agents.py
Added post_process_skill_content() hook to SkillsIntegration (identity default)
ClaudeIntegration overrides it to inject user-invocable: true and disable-model-invocation: false
Wired through presets.py and extensions.py so all skill-generation paths go through the integration

Hook command name normalization

ClaudeIntegration.post_process_skill_content() injects a dot-to-hyphen note into hook sections of generated SKILL.md files, so Claude maps speckit.git.commit → /speckit-git-commit

PowerShell encoding fix

Replaced Unicode ✓ with ASCII [OK] in both auto-commit.ps1 and auto-commit.sh

Tests

9 new tests for disable-model-invocation (positive: value is false, negative: true never appears, non-Claude agents unaffected)
5 new tests for hook note injection (positive: present in hook skills, negative: absent without hooks, idempotent, preserves indentation)
4 new tests for auto-commit scripts (positive: [OK] in output, negative: no Unicode ✓)

All 1433 tests pass.

- Set disable-model-invocation to false so Claude can invoke extension skills (e.g. speckit-git-feature) from within workflow skills - Inject dot-to-hyphen normalization note into Claude SKILL.md hook sections so the model maps extension.yml command names to skill names - Replace Unicode checkmark with ASCII [OK] in auto-commit scripts to fix PowerShell encoding errors on Windows - Move Claude-specific frontmatter injection to ClaudeIntegration via post_process_skill_content() hook on SkillsIntegration, wired through presets and extensions managers - Add positive and negative tests for all changes Fixes github#2178

Copilot

Pull request overview

Fixes Claude Code failing to execute chained speckit skills/hooks by ensuring generated skills are user-invocable, model-invocation is enabled (for skill chaining), and hook documentation clarifies dot-to-hyphen command normalization; also removes Unicode output that breaks PowerShell parsing on Windows.

Changes:

Moves Claude-specific SKILL.md frontmatter mutation into ClaudeIntegration.post_process_skill_content() and routes preset/extension skill generation through integration post-processing.
Injects a dot→hyphen normalization note into hook instruction sections in Claude-generated skills.
Replaces Unicode ✓ with ASCII [OK] in auto-commit scripts and adds/updates tests to lock in behavior.

Show a summary per file

File	Description
`src/specify_cli/agents.py`	Removes Claude-only SKILL frontmatter from shared builder to avoid affecting other skill-generation paths.
`src/specify_cli/integrations/base.py`	Adds `SkillsIntegration.post_process_skill_content()` hook (default identity) for integration-specific skill post-processing.
`src/specify_cli/integrations/claude/__init__.py`	Implements Claude-specific post-processing: injects `user-invocable: true`, `disable-model-invocation: false`, and a hook command normalization note.
`src/specify_cli/presets.py`	Applies integration post-processing to preset-created/restored SKILL.md content.
`src/specify_cli/extensions.py`	Applies integration post-processing to extension-created SKILL.md content.
`extensions/git/scripts/bash/auto-commit.sh`	Replaces Unicode checkmark success output with `[OK]` for portability.
`extensions/git/scripts/powershell/auto-commit.ps1`	Replaces Unicode checkmark success output with `[OK]` to avoid Windows PowerShell parser/encoding issues.
`tests/integrations/test_integration_claude.py`	Updates expectations for `disable-model-invocation: false` and adds coverage for Claude post-processing + hook note injection behavior.
`tests/test_extension_skills.py`	Updates extension skill YAML expectations to match new Claude post-processing (`disable-model-invocation: false`).
`tests/test_presets.py`	Updates preset skill assertions to expect `disable-model-invocation: false`.
`tests/extensions/git/test_git_extension.py`	Adds tests ensuring `[OK]` is used and Unicode `✓` is absent in script outputs.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Files reviewed: 11/11 changed files
Comments generated: 3

Copilot · 2026-04-15T13:58:19Z

src/specify_cli/integrations/claude/__init__.py

+        return re.sub(
+            r"(?m)^(\s*)(- For each executable hook, output the following)",
+            lambda m: m.group(1) + _HOOK_COMMAND_NOTE.rstrip("\n") + "\n" + m.group(0),


_inject_hook_command_note() always inserts a \n newline. Elsewhere in this module the injection helpers preserve the existing EOL style (\n vs \r\n), so this can introduce mixed line endings if a SKILL.md uses CRLF. Consider preserving the existing line-ending style when inserting the hook note for consistency.

Suggested change

return re.sub(

r"(?m)^(\s*)(- For each executable hook, output the following)",

lambda m: m.group(1) + _HOOK_COMMAND_NOTE.rstrip("\n") + "\n" + m.group(0),

def repl(m: re.Match[str]) -> str:

indent = m.group(1)

instruction = m.group(2)

eol = m.group(3)

return (

indent

+ _HOOK_COMMAND_NOTE.rstrip("\n")

+ eol

+ indent

+ instruction

+ eol

)

return re.sub(

r"(?m)^(\s*)(- For each executable hook, output the following)(\r\n|\n|$)",

repl,

Copilot · 2026-04-15T13:58:20Z

src/specify_cli/presets.py

+    @staticmethod
+    def _post_process_skill(agent_key: str, content: str) -> str:
+        """Delegate to the integration's post_process_skill_content if available."""
+        if not isinstance(agent_key, str) or not agent_key:
+            return content
+        from specify_cli.integrations import get_integration
+
+        integration = get_integration(agent_key)
+        if integration is not None and hasattr(integration, "post_process_skill_content"):
+            return integration.post_process_skill_content(content)
+        return content


_post_process_skill() is duplicated here and in specify_cli/extensions.py. This duplication increases the risk of future drift (e.g., changing integration lookup or error handling in one place but not the other). Consider extracting a shared helper (module-level function) and reusing it in both call sites.

Copilot · 2026-04-15T13:58:20Z

src/specify_cli/extensions.py

+    @staticmethod
+    def _post_process_skill(agent_key: str, content: str) -> str:
+        """Delegate to the integration's post_process_skill_content if available."""
+        if not isinstance(agent_key, str) or not agent_key:
+            return content
+        from specify_cli.integrations import get_integration
+
+        integration = get_integration(agent_key)
+        if integration is not None and hasattr(integration, "post_process_skill_content"):
+            return integration.post_process_skill_content(content)
+        return content


_post_process_skill() is duplicated here and in specify_cli/presets.py. To reduce maintenance overhead and avoid inconsistencies, consider consolidating this into a single shared helper used by both extension and preset skill generators.

mnriem · 2026-04-15T14:05:13Z

@copilot apply changes based on the comments in this thread

- Preserve line-ending style (CRLF/LF) in _inject_hook_command_note instead of always inserting \n, matching the convention used by other injection helpers in the same module. - Extract duplicated _post_process_skill() from extensions.py and presets.py into a shared post_process_skill() function in agents.py. Both modules now import and call the shared helper.

The regex in _inject_hook_command_note only matched lines ending immediately after 'output the following', but the actual template lines continue with 'based on its `optional` flag:'. Use [^\r\n]* to capture the rest of the line before the EOL.

Copilot

Pull request overview

This PR fixes Claude Code hook execution for Speckit by ensuring Claude-generated skills are invocable/chained, normalizing hook command naming expectations, and removing a Unicode character that breaks PowerShell parsing on Windows.

Changes:

Move Claude-specific SKILL.md frontmatter behavior out of the shared build_skill_frontmatter() and into Claude’s integration post-processing (sets disable-model-invocation: false, user-invocable: true).
Post-process SKILL.md content during preset/extension skill generation so Claude flags (and hook note) apply consistently across generation paths.
Replace Unicode ✓ with ASCII [OK] in auto-commit scripts and add tests to prevent regressions.

Show a summary per file

File	Description
`src/specify_cli/agents.py`	Adds `post_process_skill()` helper to route skill content through the selected integration’s post-processor.
`src/specify_cli/integrations/base.py`	Introduces `post_process_skill_content()` hook (default no-op) for skills integrations.
`src/specify_cli/integrations/claude/__init__.py`	Implements Claude-specific post-processing: injects `user-invocable: true`, `disable-model-invocation: false`, and a dot→hyphen hook command note.
`src/specify_cli/presets.py`	Applies integration post-processing to generated/restored preset SKILL.md files.
`src/specify_cli/extensions.py`	Applies integration post-processing to generated extension SKILL.md files.
`extensions/git/scripts/bash/auto-commit.sh`	Replaces Unicode checkmark with `[OK]` in success output.
`extensions/git/scripts/powershell/auto-commit.ps1`	Replaces Unicode checkmark with `[OK]` in success output (avoids Windows PowerShell parsing issues).
`tests/integrations/test_integration_claude.py`	Expands tests for Claude frontmatter flags and hook-note injection behavior.
`tests/test_extension_skills.py`	Updates expectation for Claude-generated extension SKILL.md to have `disable-model-invocation: false`.
`tests/test_presets.py`	Updates expectations for preset SKILL.md to have `disable-model-invocation: false`.
`tests/extensions/git/test_git_extension.py`	Adds tests asserting `[OK]` output and absence of Unicode checkmark for both bash and PowerShell scripts.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Files reviewed: 11/11 changed files
Comments generated: 0 new

Instead of a free function in agents.py that re-resolves the integration by key, callers in extensions.py and presets.py now resolve the integration once via get_integration() and call integration.post_process_skill_content() directly. The base identity method lives on SkillsIntegration.

Copilot AI review requested due to automatic review settings April 15, 2026 13:53

Copilot started reviewing on behalf of mnriem April 15, 2026 13:53 View session

Copilot AI reviewed Apr 15, 2026

View reviewed changes

mnriem self-assigned this Apr 15, 2026

mnriem added 2 commits April 15, 2026 09:29

Copilot AI review requested due to automatic review settings April 15, 2026 14:49

Copilot started reviewing on behalf of mnriem April 15, 2026 14:50 View session

Copilot AI reviewed Apr 15, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: allow Claude to chain skills for hook execution#2227

fix: allow Claude to chain skills for hook execution#2227
mnriem wants to merge 4 commits intogithub:mainfrom
mnriem:fix/claude-hook-chaining-2178

mnriem commented Apr 15, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 15, 2026

Uh oh!

Copilot AI Apr 15, 2026

Uh oh!

Copilot AI Apr 15, 2026

Uh oh!

mnriem commented Apr 15, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-        return re.sub(
-            r"(?m)^(\s*)(- For each executable hook, output the following)",
-            lambda m: m.group(1) + _HOOK_COMMAND_NOTE.rstrip("\n") + "\n" + m.group(0),
+        def repl(m: re.Match[str]) -> str:
+            indent = m.group(1)
+            instruction = m.group(2)
+            eol = m.group(3)
+            return (
+                indent
+                + _HOOK_COMMAND_NOTE.rstrip("\n")
+                + eol
+                + indent
+                + instruction
+                + eol
+            )
+        return re.sub(
+            r"(?m)^(\s*)(- For each executable hook, output the following)(\r\n|\n|$)",
+            repl,

Conversation

mnriem commented Apr 15, 2026

Summary

Changes

Core fix: disable-model-invocation → false

Hook command name normalization

PowerShell encoding fix

Tests

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Copilot's findings

Uh oh!

Copilot AI Apr 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 15, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 15, 2026

Choose a reason for hiding this comment

Uh oh!

mnriem commented Apr 15, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Copilot's findings

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Core fix: `disable-model-invocation` → `false`