fix(init): delegate skills scope + agent picker to skills CLI by Railly · Pull Request #114 · clerk/cli

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly · 2026-04-07T15:18:44Z

Problem

clerk init was forcing skills install to all 45 supported agents non-interactively by passing --agent '*' -y. Result: 30+ agent folders (.claude/, .cursor/, .cline/, ...) in the project root + ~25 lines added to .gitignore, even when the user only uses 1-2 agents.

Reported by @rafael-thayto in Slack.

Root cause

The skills CLI (vercel-labs/skills) already has the UX we want:

Auto-detects installed agents on the system if --agent is not passed
Shows a scope picker (project vs global) when run interactively
Has a multi-select for fine-tuning agents

We were stomping on all of that.

Fix

Drop --agent '*' -y in interactive human mode and inherit stdin so the skills CLI's pickers actually work. Non-interactive mode (agent or -y) falls back to -y -g so CI keeps working.

Mode	Flags	Result
Human (interactive)	(none)	Skills CLI auto-detects + shows scope picker
`clerk init -y`	`-y -g`	Non-interactive, global, auto-detected agents
Agent mode	`-y -g`	Same as `-y`

clerk init keeps its single opt-in prompt upfront ("Install agent skills?"). Everything past that — scope, agents — is the user's call via the skills CLI's native interface.

Tests

Extracted buildSkillsArgs(skills, interactive) as a pure function
3 unit tests covering: interactive mode passes no -y/-g/--agent, non-interactive includes -y -g, neither mode passes --agent (let auto-detect work)
Full init test suite still green (156 tests)

Closes AIE-791
Follow-up to #86

The skills CLI already auto-detects installed agents and shows a scope picker (project vs global) when run interactively. We were stomping on both by passing `--agent '*' -y`, which forced install to all 45 supported agents non-interactively and dropped the user into a project-scoped install with 30+ agent folders cluttering the repo and ~25 lines added to .gitignore. Fix: - In interactive human mode, drop both `--agent` and `-y` so the skills CLI's native UX takes over (auto-detect + scope picker + agent multi-select). - In non-interactive mode (agent mode or `clerk init -y`), pass `-y -g` so it runs unattended with global scope and only the agents the user actually has installed. - Inherit stdin so the picker actually works in interactive mode. `clerk init` keeps its single opt-in prompt upfront. Everything past that — scope, agents — is the user's call via the skills CLI's native interface. Extracted `buildSkillsArgs` as a pure function so the flag logic is unit-testable without mocking Bun.spawn. Reported by Rafa. Closes AIE-791

coderabbitai · 2026-04-07T16:00:28Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Free

Run ID: bf872f35-f2d3-4715-8b0e-5162c571c9b1

📥 Commits

Reviewing files that changed from the base of the PR and between ebdb1c9 and 4999528.

📒 Files selected for processing (2)

packages/cli-core/src/commands/init/skills.test.ts
packages/cli-core/src/commands/init/skills.ts

📝 Walkthrough

Walkthrough

A new helper function buildSkillsArgs() has been added to construct command-line arguments for the npx skills add command based on selected skills and an interactive mode flag. The installSkills function has been refactored to use this helper, conditionally including flags (-y -g) only in non-interactive mode while allowing interactive mode to rely on the skills CLI's native interface. A corresponding test file validates the function's behavior across both interactive and non-interactive scenarios, confirming that --agent is never passed.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Railly marked this pull request as ready for review April 7, 2026 16:00