Executes approved shell commands, manages backups, and safely retrieves secrets from Bitwarden.

Do not install this skill until the developer clarifies several gaps: 1) how the 'bws' Bitwarden wrapper is authenticated (what env vars or tokens are required and how they are stored), 2) what the 'exec-approvals' gate is and how it enforces/records human approvals (avoid vague 'wait for Paul' instructions), 3) whether the hard-coded path (/home/intelad/...) matches your environment or will be configurable, and 4) where logs ('memory/YYYY-MM-DD.md') are written and who can read them. Because the skill directs the agent to run shell commands and fetch secrets, verify provenance (source/homepage unknown) and prefer a version that declares required credentials and a machine-enforceable approval mechanism before granting access or allowing autonomous invocation.

Type: OpenClaw Skill Name: automation-runner Version: 1.0.0 The 'automation-runner' skill is highly suspicious due to its inherent capabilities and the nature of its security controls. The SKILL.md explicitly grants the AI agent the ability to execute arbitrary shell commands (`exec`) and retrieve secrets from Bitwarden (`bws`). While it outlines security directives like `exec-approvals` and workspace limitations, these are instructions to the agent and are highly susceptible to prompt injection. An attacker could craft a prompt to bypass the `/approve` step, execute commands outside the `/home/intelad/.openclaw/workspace/scripts` directory, or exfiltrate retrieved secrets, leading to potential RCE and data theft. This represents a critical vulnerability, not intentional malice within the skill bundle itself.