Model Handoff

Maintain a HANDOFF.md file in the workspace so context survives seamlessly when switching between LLM models (e.g. Claude → GPT → Gemini). Use when the user...

This skill appears to do what it says (create and keep a HANDOFF.md for switching models), but take precautions before installing/using it: do not store secrets or private data in HANDOFF.md; avoid referencing credential filenames unless those files are tightly access-controlled; require user confirmation before the agent updates HANDOFF.md or disable autonomous writes; limit which agents/models can read the workspace; and review HANDOFF.md changes manually in sensitive projects. If you want the skill to be safer, ask the author to (1) remove instructions that encourage pointing to credential filenames, (2) add an explicit safety rule to never include or reference secrets or private PII in any form, and (3) require explicit user approval before any proactive update to HANDOFF.md. Additional useful info to lower risk: whether the agent platform enforces file-permission isolation and whether skill-driven writes require user confirmations.

Type: OpenClaw Skill Name: model-handoff Version: 1.0.0 The 'model-handoff' skill is designed to maintain session context across different LLM models by managing a 'HANDOFF.md' file. The instructions in SKILL.md and the template in references/template.md explicitly forbid inlining secrets and focus on project status, file paths, and behavioral rules, which aligns with its stated purpose without exhibiting malicious intent or high-risk behaviors.