Epic: Add OpenClaw skill and `2kc openclaw` CLI command

[helixclaw]

Summary

Add an OpenClaw AgentSkill package to the project and a 2kc openclaw CLI command for installing/uninstalling the skill into an OpenClaw workspace.

Context

2keychains is a secret broker designed for AI agent workflows. To integrate seamlessly with OpenClaw-powered agents, we need a discoverable skill package that teaches agents how to use 2kc commands. The skill lives in the repo as skill/SKILL.md and gets symlinked into the user's OpenClaw workspace via the CLI.

Goals

1. Skill Package: A skill/ directory at project root containing SKILL.md with proper YAML frontmatter and agent instructions for using 2kc
2. CLI Command: 2kc openclaw install and 2kc openclaw uninstall subcommands that manage a symlink at ~/.openclaw/workspace/skills/2keychains
3. Prerequisite Validation: Verify OpenClaw workspace exists before attempting operations

Design Decisions

• Symlink approach: The skill directory is symlinked (not copied) so updates to SKILL.md are immediately available without reinstalling
• Workspace skills path: ~/.openclaw/workspace/skills/2keychains — workspace-level skills have highest precedence in OpenClaw's loading order
• Skill directory at project root: skill/ (not skills/) since this project provides a single skill

Dependencies

• Depends on Project bootstrap & CLI skeleton #2 (project bootstrap & CLI skeleton) — completed

Acceptance Criteria

• skill/SKILL.md exists with valid OpenClaw frontmatter and agent-facing documentation
• 2kc openclaw install symlinks skill/ to ~/.openclaw/workspace/skills/2keychains
• 2kc openclaw uninstall removes the symlink
• Both commands validate that ~/.openclaw/workspace/ exists
• Idempotent: install when already installed and uninstall when not installed are handled gracefully
• Tests cover install, uninstall, prerequisite checks, and idempotency

[helixclaw]

Architectural Decision Record

Design Rationale

Symlink over copy: The skill is symlinked into the OpenClaw workspace rather than copied. This means edits to skill/SKILL.md are immediately reflected without re-running install. This is the standard pattern for development-time skill installation.

Workspace-level placement: Skills installed to ~/.openclaw/workspace/skills/ have the highest precedence in OpenClaw's skill loading order (workspace > managed > bundled). This ensures the 2keychains skill is always discoverable when installed.

Single skill/ directory: The project provides one skill, so we use skill/ (singular) at the project root rather than a skills/ directory containing subdirectories.

Core logic separation: Filesystem operations are extracted into src/core/openclaw.ts and the CLI wiring stays in src/cli/openclaw.ts. This makes the install/uninstall logic independently testable without invoking Commander.

Key Decisions

Decision	Choice	Alternative Considered
Skill placement	~/.openclaw/workspace/skills/2keychains	~/.openclaw/skills/ (managed, lower precedence)
Install mechanism	Symlink	Copy (would require reinstall on changes)
Prerequisite check	Verify ~/.openclaw/workspace/ exists	Auto-create workspace directory
Conflict handling	Warn and refuse to overwrite	Force overwrite with --force flag

Recommended Execution Order

1. Create OpenClaw AgentSkill package (skill/SKILL.md) #12 — Create skill/SKILL.md (content-only, no code)
2. Implement 2kc openclaw install/uninstall CLI commands #13 — Implement CLI commands + tests (depends on Create OpenClaw AgentSkill package (skill/SKILL.md) #12)

Child Issues Created

Issue	Title	Dependencies
#12	Create OpenClaw AgentSkill package (skill/SKILL.md)	None
#13	Implement 2kc openclaw install/uninstall CLI commands	#12

Dependency Graph

graph TD
    12["Create OpenClaw AgentSkill package"]
    13["Implement openclaw install/uninstall commands"]
    12 --> 13
    click 12 "https://github.com/helixclaw/2keychains/issues/12"
    click 13 "https://github.com/helixclaw/2keychains/issues/13"

Loading