hindsight

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt documents a config containing an api_key and shows using it on the command line (hindsight configure --api-key ), which encourages embedding/verbatim use of secrets in commands or generated output and thus creates exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.65). Outsider free text can enter the LLM context via the OpenClaw hindsight-memory plugin’s before_prompt_build hook, which auto-recalls memories that were previously captured from message_received (i.e., user-authored chat messages on authenticated channels that the operating user did not author/choose to introduce), and injects those recalled memory texts into the agent prompt.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's setup instructions explicitly fetch and execute remote code (high-confidence examples: "curl -fsSL https://hindsight.vectorize.io/get-cli | bash" and pulling/running the remote container image ghcr.io/vectorize-io/hindsight:latest or cloning repos like https://github.com/anthropics/hindsight.git), which are runtime external dependencies that execute remote code and are required for the skill to run.