OC Evolver Restart Loop Fix
v0.1.0Repair Evolver restart storms caused by singleton lock/PID false positives and service restart policy mismatch.
⭐ 0· 292·2 current·2 all-time
by@xyezir
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description (fixing Evolver restart storms due to stale locks/PID false positives and restart-policy mismatches) align with the runtime instructions (inspect service state, check lock/PID, clear stale lock, patch lock guard, restart service). The actions requested are what an on‑host troubleshooting/fix skill would legitimately do.
Instruction Scope
SKILL.md is high-level and does not list specific commands, file paths, or service manager (e.g., systemd vs. upstart). It instructs the agent to inspect/modify lock files and patch a lock guard — operations that typically require elevated privileges and can be destructive if done incorrectly. The ambiguity grants broad discretion to the agent; require explicit, narrow steps or human approval before making changes.
Install Mechanism
Instruction-only skill with no install spec or downloaded artifacts. Nothing will be written to disk by installation, which minimizes installation risk.
Credentials
The skill declares no required environment variables, credentials, or config paths. That is proportionate to the stated on-host troubleshooting purpose.
Persistence & Privilege
always is false and the skill does not request permanent presence or modify other skills or global agent settings. Autonomous invocation is allowed by platform default but is not combined with other broad privileges here.
Assessment
This skill appears to do what it claims, but the instructions are intentionally vague and imply privileged on-host changes (clearing lock files, patching guard logic, restarting services). Before installing or letting an agent run this skill: 1) require the agent to present exact commands and file paths it will run and get explicit human approval for each privileged step; 2) run fixes first in a staging/test environment or take backups/snapshots; 3) constrain the scope by specifying the service manager (systemd/upstart), service name, and exact lock/PID file locations; 4) ask the skill (or author) for a sample patch diff and verification steps; and 5) ensure any external artifacts redact infrastructure/identity details as recommended. If you don't have a retained, trusted human operator to vet changes, treat this as higher risk.Like a lobster shell, security has layers — review code before you run it.
latestvk979x4p4hqdhaf1smevs7vd2ds822acm
License
MIT-0
Free to use, modify, and redistribute. No attribution required.