fluid-memory

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). Insecure: the prompt tells the LLM to embed the user's original text verbatim into tool calls (and shows a CLI invocation using --content "..."), so any user-supplied secrets would be echoed, stored in plaintext, and may be exposed via command-line arguments or logs.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). Both links point to an untrusted/unknown distribution channel (clawhub.ai) and a GitHub repo from an unfamiliar user that contains installable scripts which write files, spawn subprocesses, and run maintenance — making them potentially unsafe to download and execute without review.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). Hook fluid-memory-sync listens to runtime event message:sent and appends event.content (the other party’s message text, i.e., outsider-authored chat content) into ~/.openclaw/workspace/database/conversation_log.txt, which is later read/processed during the flush-triggered fluid_increment_summarize path.