Workspace Standard

Set up and maintain a structured OpenClaw workspace with project boundaries, role-based file taxonomy, and memory budgets. Use when: (1) bootstrapping a new...

This skill appears coherent and limited to local workspace organization. Before installing or running the scripts: (1) review the two shell scripts to confirm you understand what they create; (2) run them in a git-tracked workspace (so you can inspect and revert changes with git); (3) avoid using --force unless you intend to overwrite templates; (4) if you choose the README's curl/git install routes, verify the URLs are correct (they point to GitHub raw content) before executing; (5) remember the agent may autonomously consult this skill when deciding where to write files — if you prefer to control changes manually, avoid granting the agent unrestricted autonomous actions.

Type: OpenClaw Skill Name: workspace-standard Version: 1.0.0 The skill bundle contains shell injection vulnerabilities in `scripts/workspace-audit.sh` and `scripts/workspace-init.sh`. In `workspace-init.sh`, command-line arguments like `--project NAME` and `--path DIR` are directly used in `mkdir -p` and `echo` commands without sanitization, allowing arbitrary command execution if `NAME` or `DIR` contain shell metacharacters. Similarly, `workspace-audit.sh` uses values extracted from `.workspace-standard.yml` (e.g., `PROJ_SUBDIRS`) and markdown front-matter (e.g., `updated` date) directly in shell commands (`for` loops, `date -d`), creating potential RCE vectors if these inputs are controlled by an attacker. While the skill's stated purpose is benign workspace management, these vulnerabilities could be exploited to execute arbitrary commands on the host system.