signals-scout-feature-flags

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt explicitly instructs the agent to quote event-supplied feature-flag keys in reports and scratchpad entries (even if truncated/sanitized), which forces the LLM to reproduce untrusted strings verbatim and therefore can exfiltrate secret-looking values present in events.

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow ingests outsider-authored free text from event payload properties—specifically $feature_flag and $feature_flag_response inside execute-sql queries over events (including the “ghost flags” anti-join where arbitrary key strings are surfaced and treated as data at runtime).