ask-many-models

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires including "actual excerpts, code snippets, or data" verbatim in prompts and shows passing the full prompt inline to a CLI command, which can force the agent to read and emit secret values (and even place them on a command line) — an insecure credential-handling pattern.

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

• Third-party content exposure detected (medium risk: 0.30). At runtime, outsider-authored free text from the user’s prompt (and optional system prompt text) is directly embedded into the query prompt that is sent to selected models for generation and synthesis (including deep-research inputs), via commands/amm.md → SKILL.md Step 1/1b/1d/2/4 and then scripts/query.ts (fullPrompt/systemPrompt → generateText/queryDeepResearch).