Handover Hangover
v1.1.3Seamless model handoff for OpenClaw agents. Ensures continuity when the gateway switches to a fallback model mid-session.
⭐ 0· 47·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (handoff continuity) align with requested binaries (bash), files written (memory/handoff-note.md, current-task.md, .handoff-*), and the included watchdog script. The script's workspace file listing and git-status checks are proportionate to producing a mechanical fallback note.
Instruction Scope
SKILL.md confines behavior to writing/reading handoff notes and checking a small set of memory files; it does not instruct network calls or exfiltration. It does direct models to write structured notes and to check for .handoff-pending. Note: the protocol asks the model to inspect workspace context (current-task.md) and the script scans recent workspace files and git status — expected for the purpose but worth reviewing if sensitive files exist in the workspace.
Install Mechanism
Instruction-only skill plus a local bash script; no external downloads or package installs. No high-risk install mechanisms are present.
Credentials
No credentials or secret environment variables required. The script optionally uses WORKSPACE (defaulted) and standard CLI tools (find, git). The environment access requested is minimal and consistent with producing filesystem-based handoff notes.
Persistence & Privilege
The skill is declared always: true (prompt injected every turn). This is justified by the skill's goal (per-turn detection and write-side protocol) but is a notable privilege: it means the system prompt content is loaded into every agent turn. The watchdog script itself is not auto-executed by OpenClaw; the README explains explicit wiring (afterTurn/boot/heartbeat).
Assessment
This skill appears to do what it says: it provides a prompt-layer protocol and a local bash watchdog to ensure a handoff note exists for incoming models. Before enabling it, review and accept these points: (1) always:true means the SKILL.md prompt is injected into every agent turn — review the prompt content to ensure you are comfortable with that system-level instruction being present always; (2) the included script reads the workspace for recent filenames and git status and writes memory/handoff-note.prev.md, memory/.handoff-pending, and memory/.handoff-sealed — ensure no sensitive files you mind being referenced by name are placed in the workspace; (3) OpenClaw does not auto-run scripts: wiring the watchdog into afterTurn/boot/heartbeat is a manual step — evaluate whether you want that extra automation; (4) test the skill in a safe/isolated workspace first (or with a copy of your workspace) to observe generated notes and seal behavior; (5) if you accept the always-on prompt, you can proceed, otherwise consider using the script manually or gating the skill to non-always usage. If you want, I can highlight exact lines in SKILL.md or scripts/handoff.sh that reference workspace paths, file patterns, or the signal lifecycle so you can review them more closely.Like a lobster shell, security has layers — review code before you run it.
Plugin bundle (nix)
Skill pack · CLI binary · Config
SKILL.mdCLIConfig
Config requirements
State dirsmemory/
fallbackvk977qydfc7d2nvsvmvc8k29p5184c7k8handoffvk977qydfc7d2nvsvmvc8k29p5184c7k8latestvk977qydfc7d2nvsvmvc8k29p5184c7k8model-switchvk977qydfc7d2nvsvmvc8k29p5184c7k8systemvk977qydfc7d2nvsvmvc8k29p5184c7k8
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🔄 Clawdis
Binsbash