Back to skill
Skillv0.1.1
ClawScan security
Managed Agent Growth Loop · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignApr 30, 2026, 12:54 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill is an instruction-only operational playbook for improving managed agents and its requests are proportionate to that purpose, with only minor notes about implicit platform dependencies.
- Guidance
- This skill is an operational playbook (state templates, watchlists, cron recipes) and is internally coherent. Before installing, confirm two things: (1) your OpenClaw runtime already provides the openclaw CLI and configured channels/integrations (Discord/webhook tokens, etc.) referenced in the examples — the skill does not declare or request those credentials; (2) where the skill suggests storing state (memory/clients/...) is an appropriate, secure place in your environment. Also review any cron jobs you create to ensure they only draft/monitor (not automatically send) unless you explicitly approve external deliveries.
Review Dimensions
- Purpose & Capability
- noteThe name/description (agent operational growth loop) matches the instructions (usage maps, state handoff, watchlists, cron recipes, review flow). One minor inconsistency: the SKILL.md shows example openclaw CLI commands and recommends storing state under memory/clients/... which implies the agent runtime provides an openclaw binary and writable memory paths, but the skill's metadata does not declare any required binaries/config paths. This is plausible (examples assume platform tools) but should be confirmed.
- Instruction Scope
- okAll runtime instructions stay within the stated purpose: categorize usage, create minimal state handoffs, maintain watchlists, schedule periodic checks, and perform multi-agent review. The doc explicitly limits actions that send external messages (user approval required for sensitive sends) and frames cron jobs as monitoring/draft-generators rather than direct outbound senders.
- Install Mechanism
- okThere is no install spec and no code files — this is instruction-only, so nothing will be written to disk by the skill itself. Low install risk.
- Credentials
- noteThe skill declares no required environment variables or credentials, which is consistent with being an instruction-only playbook. However, the examples reference posting/announcing via channels (e.g., --channel discord --to "channel:{운영채널ID}") and using openclaw cron; those actions depend on preconfigured platform integrations/credentials and writable agent memory. Users should verify that any channel/webhook credentials and the openclaw CLI are already provisioned by the platform — the skill does not request nor explain them.
- Persistence & Privilege
- okalways:false and user-invocable:true. The skill suggests creating cron/heartbeat jobs via platform CLI (examples) but does not itself claim forced inclusion or modification of other skills. This is proportionate for an ops/playbook skill.