The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local TypeScript scripts (discover-openclaw.ts, extract-channel-credentials.ts) using pnpm exec tsx to automate discovery and extraction tasks.
The execution is scoped to the skill's own script directory and existing project tools (ncl, onecli).
Phase 8 involves running build and test commands (pnpm run build, vitest) to validate the migration logic against the host environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests configuration and workspace data from a legacy OpenClaw installation, which represents a potential attack surface if the legacy data contains malicious instructions.
Ingestion points: Data is read from openclaw.json, clawdbot.json, and various workspace files like IDENTITY.md and SOUL.md (documented in SKILL.md and discover-openclaw.ts).
Boundary markers: The skill explicitly requires a conversational, phase-based migration where the agent must summarize findings and ask for user confirmation before applying any changes ("Never silently copy data. Read it, explain it, place it, then apply").
Capability inventory: The agent has access to file system operations (cp, rm, sed), shell command execution, and credential management via the OneCLI Vault.
Sanitization: Sensitive values are programmatically masked using the maskCredential function in transform.ts before being displayed in the chat transcript, and the routing logic for secrets (Vault vs. .env) is handled by deterministic code rather than LLM discretion.

migrate-from-openclaw