Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed tariff lookup client that sends product details to an external tariff API, with no evidence of hidden persistence, credential access, or destructive behavior.
Before installing, treat Accio/TurtleClassify as a third-party service: only submit product names and metadata you are comfortable sharing externally, avoid unnecessary identifiers or sensitive sourcing details, and choose CSV output paths deliberately to avoid overwriting local files.
The skill clearly instructs the agent to make outbound API requests to TurtleClassify, but the manifest declares no explicit tool scope such as network permissions or allowed-tools. This creates an authorization ambiguity where a host agent may permit broader network access than intended or fail to enforce least privilege, increasing the chance of unintended external communication.
The workflow directs users to send product records to an external API but does not clearly warn that product names and trade metadata will leave the local environment. In trade and sourcing contexts, product descriptions, origins, destinations, and IDs may be commercially sensitive, so silent transmission to a third party can cause confidentiality and compliance issues.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
if opt in prod:
payload[opt] = prod[opt]
try:
resp = requests.post(endpoint, json=payload, timeout=self.timeout)
resp.raise_for_status()
outer = resp.json()
if not outer.get("success"):
The code transmits user-supplied product, country, and potentially additional product metadata to a third-party service at accio.com, but the skill itself provides no explicit user-facing disclosure, consent flow, or data minimization guarantees. In an agent skill context, this is risky because users may reasonably assume classification is local or not realize commercial/sourcing data is being sent externally.
The requests.post call sends collected product details to an external endpoint, which creates a real data exfiltration path outside the local agent environment. While this appears functionally necessary for the tariff lookup skill, it is still a security-relevant behavior because product descriptions, origin/destination data, and optional metadata may contain sensitive business information.
if key in product and product[key]:
payload[key] = product[key]
try:
r = requests.post(endpoint, json=payload, timeout=self.timeout)
r.raise_for_status()
res = r.json()
if not res.get('success'):
The usage example shows loading products.csv and writing enriched results to products_with_tariffs.csv, which affects user data on disk. The skill description does not include any warning or note that it will create or overwrite output files when following this example.
The sample output includes tariffFormula in Chinese (一般关税...附加关税...) while the rest of the documentation is in English. This may indicate the skill returns or assumes a fixed language without offering user choice or explaining the locale constraint.
No suspicious patterns detected.