extract-methodology

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md instructs the runtime workflow to run scripts/parse_sections.py on paper_text.md (a paper’s full text provided by the user/workflow), and that script reads the input file contents and outputs extracted methodology/results text into sections.json, which the agent then “evaluates … secara mendalam” for extraction—so if the paper text originates from an outsider-authored source (e.g., fetched from web/arXiv/OpenAlex/PubMed), that free-form outsider prose is ingested into the LLM context via the extracted sections.