Dreaming

Creative exploration during quiet hours. Turns idle heartbeat time into freeform thinking — hypotheticals, future scenarios, reflections, unexpected connecti...

Install only if you want an opt-in heartbeat routine that writes reflective notes to the workspace. Before enabling it, fix or constrain dreamChance handling, keep data/dream-state.json writable only by trusted components, and avoid letting the agent store sensitive private or client information in dream files.

Type: OpenClaw Skill Name: dreaming Version: 1.0.2 The skill 'dreaming' is designed for creative exploration during quiet hours, writing outputs to designated files within its workspace. It uses standard binaries (`jq`, `python3`) for state management and random chance calculation. All file operations are confined to the skill's `data/` and `memory/` directories. The `SKILL.md` instructions guide the agent to parse a `category:prompt` string and write a thoughtful exploration, which aligns with its stated purpose and does not contain any instructions for malicious actions like data exfiltration, unauthorized command execution, or prompt injection with harmful objectives. The use of `jq` with `--arg` and `--argjson` for state updates, and `jq -r` for reading, appears robust against injection. The `python3 -c` command is also not vulnerable to shell injection from the `$DREAM_CHANCE` variable.