skill-governance

OpenClaw Cognitive Operating & Skill Governance Kernel

This skill is conceptually coherent with a governance kernel, but it contains ambiguous operational directives that could change files, archive or delete other skills, and push summaries externally without specifying where. Before installing: 1) Ask the author to clarify exact filesystem paths and required permissions for /memory and archived_skills/, and which service/endpoints (and credentials) are used for 'external synchronization'. 2) Confirm which bundles the skill may mount/unmount and get explicit allow-listing for those bundles. 3) Run the skill in a restricted sandbox with monitoring to observe file writes and outgoing connections. 4) If you cannot get clear answers, avoid enabling autonomous invocation or limit its scope and privileges.

Type: OpenClaw Skill Name: skill-governance Version: 2.1.0 The skill `skill-governance` defines operational protocols for an AI agent. It is classified as suspicious due to two main capabilities outlined in `SKILL.md`: the automatic mounting of other skill bundles (e.g., `research.bundle`, `automation.bundle`) based on keywords, which poses a risk if those bundles are unvetted or malicious; and the mandatory external synchronization of task summaries for critical tasks, which is a data exfiltration capability that could be abused if the synchronization mechanism is insecure or unauthorized. These features represent significant vulnerability risks rather than explicit malicious intent.