publish

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt requires the agent to accept or generate a password and then substitute it verbatim into curl commands via the {password_header} placeholder (e.g. -H "X-Password: "), which forces the LLM to handle and embed secret values directly in generated commands.

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

• Third-party content exposure detected (medium risk: 0.30). In brewdoc:publish, the workflow ingests outsider-authored free-form text from Position 1 of $ARGUMENTS (including prose-resolved inline content) and then embeds it into the publishing payload (CONTENT=$(cat ... {content} ...)), which the LLM then sends to POST /api/html?format=markdown at runtime.