Memory Mesh Core

Builds a reusable, scored memory mesh with safety gating and 12-hour auto-refresh for cross-session memory consolidation and quality control in OpenClaw.

This skill appears to do what it claims, but it also performs powerful operations that you should review before installing. Key actions to take before use: - Audit the scripts (especially global_memory_sync.py, ensure_openclaw_cron.py, install_bootstrap.py) to confirm you accept: (a) running clawhub/openclaw/gh on your host, (b) automatic installation/updates of other skills, and (c) creation/editing of OpenClaw cron jobs. - Run in an isolated or sandbox workspace first to observe behavior and outputs (the skill reads workspace files and writes memory/memory_mesh/* artifacts). - If you will allow GitHub posting, ensure your gh credentials have only the scopes you intend and consider keeping automated posting disabled (do not pass --post-issue-comments or set setup_12h.sh posting flag to off). - If you are uncomfortable with automatic skill updates, set auto_update_skills to false in skills/memory-mesh-core/config/global_sync.json or avoid running the install_bootstrap/global sync scripts. - Verify that clawhub/openclaw/gh CLIs are from trusted sources on your machine and that you consent to them being invoked by the skill. - Consider limiting network exposure and reviewing promoted JSON outputs (memory/memory_mesh/feeds and github_issue_batch_v1.json) before any automatic posting. If you want, I can point out the exact lines or functions in the scripts that perform each privileged action, or suggest minimal configuration changes to reduce risk (eg. disable auto-update, disable scheduled posting).

Type: OpenClaw Skill Name: memory-mesh-core Version: 1.0.6 The skill is classified as suspicious due to critical prompt/shell injection vulnerabilities. The `scripts/ensure_openclaw_cron.py` and `scripts/post_global_comment_via_openclaw.py` scripts directly interpolate user-controlled arguments (e.g., `--issue-url`, `--skill-url`) into the `message` argument of `openclaw cron add/edit` commands without proper sanitization. This allows an attacker to inject arbitrary commands or malicious prompts into the scheduled tasks, leading to potential Remote Code Execution (RCE) by the OpenClaw agent. While the skill includes positive security features like secret/PII detection and output sanitization, this injection flaw represents a severe vulnerability.