ll-skill-install

name	ll-skill-install
description	Install a canonical workflow skill into Codex or Claude Code as a workspace-bound adapter that points back to the canonical implementation in the current repository. Use when LL/LEE workflow skills should be packaged or refreshed for local agent runtimes instead of hand-copying folders and manually rewriting installed SKILL.md and agents/openai.yaml.

LL Skill Install

Install canonical workflow skills into local runtime skill directories as workspace-bound adapters.

Use This Skill For

• Installing a new local workflow skill into Codex or Claude Code.
• Reinstalling an existing installed skill after the canonical source changed.
• Converting a canonical governed skill into the adapter-style installed copy used by local runtimes.

Install Workflow

1. Resolve the canonical source skill folder in the current workspace.
2. Confirm the canonical source is already the correct abstraction under E:\ai\LEE-Lite-skill-first\ssot\adr\ADR-038-运行时核心抽象边界与对象分层基线.MD. The installer packages an existing Skill authority; it does not promote a Command, Tool, workflow residue, or repo CLI carrier into a skill.
3. Run scripts/install_adapter.py with the source path.
4. Set --runtime codex or --runtime claude unless --dest-root is being passed explicitly.
5. Use --replace when refreshing an existing installed copy.
6. Validate the installed result by checking the generated SKILL.md and agents/openai.yaml.

Default command pattern:

python scripts/install_adapter.py --source <path-to-canonical-skill> --runtime codex --replace

Claude Code example:

python scripts/install_adapter.py --source <path-to-canonical-skill> --runtime claude --replace

Script Behavior

The installer script:

• copies the canonical skill into the selected runtime skills directory
• rewrites the installed SKILL.md into adapter-style wording
• rewrites agents/openai.yaml for adapter-style UI metadata
• preserves the copied governance pack, scripts, and contracts

Guardrails

• Treat the workspace skill as the canonical source of truth.
• Treat ADR-038 as the abstraction boundary baseline. Installation preserves authority and carrier boundaries; it does not reclassify runtime objects.
• Do not hand-edit the installed copy when the same change belongs in the canonical source.
• Use --replace only when you intend to refresh the installed copy.
• Prefer absolute paths for --source, --dest-root, and --workspace-root.

Resources

• scripts/install_adapter.py: installs or refreshes a workspace-bound adapter copy in Codex or Claude Code skills.

References

• E:\ai\LEE-Lite-skill-first\ssot\adr\ADR-038-运行时核心抽象边界与对象分层基线.MD
• E:\ai\LEE-Lite-skill-first\ssot\adr\ADR-020-标准 Skill 实现、Adapter 安装与 CLI 边界基线.MD