Caches and manages frequently used context fragments to avoid redundant processing by saving, listing, and clearing cached data.

This skill is coherent and simple, but review these before installing: (1) The runtime uses $CONTENT (not declared); ensure the agent will supply only intended fragments. (2) Anything appended is stored plaintext at ~/.openclaw/context-cache/fragments.md — do not cache secrets or sensitive data. (3) Consider changing the append/clear commands to safer variants (explicit file creation, use a directory listing pattern instead of aggressive rm -rf, and set file permissions). (4) If you plan to preload cached fragments on startup, audit those fragments first to avoid unintentionally exposing private information to agents. If you want stronger safety, require explicit user confirmation before memoizing and add sanitization/limits on fragment size and content.

Type: OpenClaw Skill Name: context-memoize Version: 0.1.0 The skill implements a context caching mechanism using shell commands in SKILL.md that are vulnerable to shell injection. Specifically, the use of echo "$CONTENT" >> ~/.openclaw/context-cache/fragments.md allows for arbitrary command execution if the content variable contains shell metacharacters. While the stated purpose of local context memoization appears legitimate, the lack of input sanitization is a significant security flaw.