Signal Dreaming

Signal-driven memory consolidation for OpenClaw agents. Automatically consolidates recent session logs into long-term memory using recall frequency signals f...

This skill appears internally consistent and constrained to workspace memory maintenance, but take these practical precautions before enabling it: 1) Run it first in a test workspace or a copy of your memory directory to verify behavior and outputs. 2) Review backups under <WORKSPACE_ROOT>/.backup/memory-dreams/ after a run; the skill backs up files before modifying them. 3) The included references/dream-audit.sh performs lightweight secret-pattern checks but is not a full DLP — if your workspace contains unusual credential formats or very sensitive secrets, perform an independent audit or extend the scanner. 4) Because the skill writes L2 files, MEMORY.md, and dream-log.md, ensure the provided workspace root is correct and does not point at a repo or folder containing third‑party secrets you don't want touched. 5) If you plan automated runs (cron/agentTurn), limit scheduling to trusted, isolated sessions and review the skill's diffs/outputs regularly. Overall, nothing in the bundle looks disproportionate to its stated purpose.

Type: OpenClaw Skill Name: signal-dreaming Version: 1.3.1 The signal-dreaming skill is a memory consolidation tool designed to organize agent logs based on recall frequency. It demonstrates high security awareness by including a dedicated shell script (references/dream-audit.sh) that uses regex to detect and prevent the accidental promotion of secrets (API keys, private keys) into long-term memory. The protocol (references/dream-protocol.md) enforces a three-phase safety model with mandatory backups, read-only planning phases, and explicit instructions to redact sensitive data rather than exfiltrate it.