gov-overseas-trip-report

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

• Third-party content exposure detected (medium risk: 0.30). Gov Overseas Trip Report ingests outsider-authored free text from provider pages/attachments at runtime, e.g., it fetches and parses “open_portal” and later downloads/kordoc-extracts PDF/HWP/HWPX report text from public boards like nec (the text content is authored by third parties and then fed into the LLM for summarization/feature extraction).

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I scanned the skill and fixture files for literal high-entropy credentials and ignored obvious placeholders and low-entropy example values (IDs, page indices, filenames, phone numbers, etc.).

Findings:

• tests/fixtures/open-portal.html: line 29 contains a long base64-like string assigned to wsab_licenseData. This is high-entropy, not an obvious placeholder, and looks like an embedded license/key blob — flag as a secret.
• tests/fixtures/ggc-list.html: line 138 contains Kakao.init("f79145ba0eefe2bbc9ced5dfa994e2f3"); with the inline comment "//javascript key". This is a real-looking API/javascript key (high-entropy hex), not a placeholder — flag as a secret.

Ignored items: numeric IDs (bcIdx, bbsSn, list_no), filenames, pagination parameters, and normal text/URLs — these are low-entropy or documentation data and not flagged.