The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection via data ingested from the Temu API.
Ingestion points: External data enters the agent context through API responses processed in scripts/_temu_common.py and various API-specific scripts such as scripts/eu_ads_searchrec_ad_log_query.py and scripts/eu_ads_searchrec_ad_reports_mall_query.py.
Boundary markers: There are no explicit boundary markers or LLM delimiters separating raw text data (such as seller names, change information descriptions, or campaign parameters) from structural agent instructions.
Capability inventory: The skill possesses network communication capabilities (call_temu_api via urllib.request.urlopen and scripts/onboarding.py via requests.post) and local filesystem write capabilities (emit_result writing JSON response files to <cwd>/linkfox/<date>/<session>/data/).
Sanitization: The scripts use standard json.loads to handle responses, but lack natural language validation or escaping to ensure that malicious instructions hidden within Temu ad logs or campaign names are not parsed as commands by the underlying model.

linkfox-temu-ads-eu