OpenClaw Reliability Pack
v1.0.1Unified reliability triage for OpenClaw incidents from first symptom to ship/rollback decision. Use for channel failures, gateway/runtime instability, cron/s...
⭐ 0· 91·0 current·0 all-time
byVasiliy@vassiliylakhonin
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description (reliability triage) aligns with the instructions: selecting a path (channel/core/change), running triage commands, doing baseline checks, and producing an escalation payload. The actions described are what a reliability pack would reasonably do.
Instruction Scope
SKILL.md tells the agent to run specific triage commands (openclaw-channel-doctor, openclaw-troubleshooter-pro, agent-regression-guard) and to include commands run and logs/errors in outputs. This is appropriate for triage, but the document is vague about which logs/paths to read, how to redact sensitive fields, and what external endpoints (if any) results are sent to. That vagueness could lead to overly broad reads or disclosure unless runtime policies limit it.
Install Mechanism
Instruction-only skill with no install spec and no code files — lowest-risk delivery. Nothing is downloaded or written to disk by the skill definition itself.
Credentials
The skill declares no required environment variables, credentials, or config paths. Its needs (running local triage tools and reading logs) are proportionate to a reliability triage role.
Persistence & Privilege
always is false and the skill does not request persistent/system-wide configuration or extra privileges. It does not assert autonomous permanent presence beyond normal agent invocation.
Assessment
This instruction-only skill appears coherent for incident triage, but before installing: (1) verify the referenced tools (openclaw-channel-doctor, openclaw-troubleshooter-pro, agent-regression-guard) exist and come from trusted sources; (2) restrict the agent's ability to exfiltrate logs or other data (network egress controls) and require explicit approval before sending sensitive logs off-host; (3) define/ enforce a redaction policy for logs (what to redact) because the instructions require including logs/errors but don't say how to sanitize them; (4) test the skill in a least-privilege staging environment so you can observe which commands it runs and which files it reads. These steps reduce the practical risk from the skill's intentionally broad/underspecified runtime instructions.Like a lobster shell, security has layers — review code before you run it.
agent-regression-guardvk976kfvesye1x5ap8ke4v5jemn84k0dcchannel-doctorvk976kfvesye1x5ap8ke4v5jemn84k0dclatestvk97eav0qfav1zb5kykqknc129184qm1tlatestmust-havevk976kfvesye1x5ap8ke4v5jemn84k0dcreliability-packvk976kfvesye1x5ap8ke4v5jemn84k0dctroubleshootervk976kfvesye1x5ap8ke4v5jemn84k0dc
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🛡️ Clawdis
OSLinux · macOS · Windows