social-content-parser

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

• Third-party content exposure detected (high risk: 0.85). At runtime, the skill fetches outsider-authored social post text/media via public BugPk endpoints (e.g., fetchJson() → fetchPrimaryOrFallback() → fetchWithRetry()), then writes payload/normalized.desc/normalized.title into report.md and normalized.json, which becomes LLM-readable context for downstream use.

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

• Secret detected (high risk: 1.00). I scanned the skill content for literal high-entropy credentials. I found a high-entropy query parameter value in the captured share URLs: the xsec_token parameter (e.g., "xsec_token=CBBNYcvAXoAYggeVbp3T9pc-v8xm20ckKjNF2DSiPqyhE="). This looks like a real token/signature (random-looking base64-like string) embedded in the share URL and therefore could be a usable credential or signature for the Xiaohongshu API/edge — so I flag it.

I did not flag other URL parameters (share_id, shareRedId, apptime) because those appear to be share identifiers or timestamps rather than authentication secrets. No API keys, private keys, or other obvious high-entropy secrets (e.g., "sk-..." style keys or PEM blocks) were present elsewhere in the files.