context-guardian

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill instructs the agent to extract "todas as informacoes extraidas" (including commands that worked and P0 critical items) and persist them into snapshots and MEMORY.md, which would require the LLM to include any secret values present in the session verbatim in generated outputs/files.

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes context-guardian extracting categories by “Percorrer toda a conversa” during the session, so outsider-authored free text (the user/attacker’s chat messages) is ingested by the LLM without selecting a specific item first.