You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Convert hve-core-installer.agent.md from a custom agent into a self-contained Skill package under .github/skills/hve-core-installer/. The installer's interactive workflow, environment detection scripts, validation scripts, and installation method logic should be extracted into the skill format while preserving all existing functionality.
Motivation
The hve-core-installer contains substantial domain-specific knowledge (environment detection, installation methods, validation, MCP configuration, agent copy workflows) that fits the Skill pattern better than an agent. Skills are reusable knowledge packages that can be invoked by any agent, making the installer's capabilities more broadly accessible and easier to maintain.
Acceptance Criteria
1. Extract into Skill using /prompt-build
Use /prompt-build to extract the scripts and instructions from hve-core-installer.agent.md into a .github/skills/hve-core-installer/ Skill package
Create SKILL.md with proper frontmatter (name, description) following the existing video-to-gif skill as reference
Extract all embedded scripts (environment detection, validation, extension validation, collision detection, agent copy, upgrade detection, file status check, eject) into standalone script files under .github/skills/hve-core-installer/scripts/
Provide both Bash and PowerShell variants for each script where the agent currently includes both
SKILL.md must document all 7 phases, decision matrix, MCP configuration templates, error recovery, and rollback procedures
2. Functional parity with current codebase
Verify all 6 clone-based installation methods work correctly (peer clone, git-ignored, mounted, codespaces, multi-root workspace, submodule)
Verify extension quick-install workflow works correctly
Ensure MCP configuration templates match current .vscode/mcp.json format
Ensure settings configuration templates match current chat.*FilesLocations VS Code settings format
Validate scripts reference correct paths relative to the new skill location
3. Remove the agent file
Delete .github/agents/hve-core-installer.agent.md
Remove the hve-core-installer.agent.md entry from collections/hve-core-all.collection.yml
4. Collection configuration
Create a new collections/hve-core-installer.collection.yml that includes only the hve-core-installer skill
Create a corresponding collections/hve-core-installer.collection.md describing the collection
Add the hve-core-installer skill to collections/hve-core-all.collection.yml (replacing the removed agent entry)
Do NOT add hve-core-installer to any other existing collection YAML files
Run npm run plugin:generate to regenerate plugin outputs
Run npm run plugin:validate to confirm collection metadata is correct
Implementation Notes
The existing video-to-gif skill at .github/skills/video-to-gif/ serves as the reference implementation for skill structure
Scripts should be organized under .github/skills/hve-core-installer/scripts/ with clear naming (e.g., detect-environment.sh, detect-environment.ps1, validate-installation.sh, validate-installation.ps1)
The SKILL.md should reference scripts using relative paths from the skill directory
All interactive prompts, checkpoint text, and decision matrix tables from the agent should be preserved in the SKILL.md
The agent's tool declarations (vscode/newWorkspace, vscode/runCommand, etc.) are not applicable to skills and should be omitted
Summary
Convert
hve-core-installer.agent.mdfrom a custom agent into a self-contained Skill package under.github/skills/hve-core-installer/. The installer's interactive workflow, environment detection scripts, validation scripts, and installation method logic should be extracted into the skill format while preserving all existing functionality.Motivation
The hve-core-installer contains substantial domain-specific knowledge (environment detection, installation methods, validation, MCP configuration, agent copy workflows) that fits the Skill pattern better than an agent. Skills are reusable knowledge packages that can be invoked by any agent, making the installer's capabilities more broadly accessible and easier to maintain.
Acceptance Criteria
1. Extract into Skill using
/prompt-build/prompt-buildto extract the scripts and instructions fromhve-core-installer.agent.mdinto a.github/skills/hve-core-installer/Skill packageSKILL.mdwith proper frontmatter (name,description) following the existingvideo-to-gifskill as reference.github/skills/hve-core-installer/scripts/SKILL.mdmust document all 7 phases, decision matrix, MCP configuration templates, error recovery, and rollback procedures2. Functional parity with current codebase
.vscode/mcp.jsonformatchat.*FilesLocationsVS Code settings format3. Remove the agent file
.github/agents/hve-core-installer.agent.mdhve-core-installer.agent.mdentry fromcollections/hve-core-all.collection.yml4. Collection configuration
collections/hve-core-installer.collection.ymlthat includes only thehve-core-installerskillcollections/hve-core-installer.collection.mddescribing the collectionhve-core-installerskill tocollections/hve-core-all.collection.yml(replacing the removed agent entry)hve-core-installerto any other existing collection YAML filesnpm run plugin:generateto regenerate plugin outputsnpm run plugin:validateto confirm collection metadata is correctImplementation Notes
video-to-gifskill at.github/skills/video-to-gif/serves as the reference implementation for skill structure.github/skills/hve-core-installer/scripts/with clear naming (e.g.,detect-environment.sh,detect-environment.ps1,validate-installation.sh,validate-installation.ps1)SKILL.mdshould reference scripts using relative paths from the skill directorySKILL.mdvscode/newWorkspace,vscode/runCommand, etc.) are not applicable to skills and should be omittedRelated Files
.github/agents/hve-core-installer.agent.md.github/skills/video-to-gif/SKILL.mdcollections/hve-core-all.collection.ymldocs/getting-started/methods/