magic-linguistic-speech

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

• Third-party content exposure detected (high risk: 0.78). The skill’s runtime workflow ingests outsider-authored free text from user-provided annotation files (e.g., ELAN EAF tier labels / Praat TextGrid interval text / FLEx XML exported text / SayMore IMDI-extracted ELAN content) into the LLM context via the Step 1–5 “Speech Plan” handoff and tier/text parsing paths.