Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed Jina AI integration that sends user-provided URLs, search queries, research prompts, and a Jina API key to Jina endpoints for web reading, search, and DeepSearch.
Install only if you are comfortable sending requested URLs, search terms, research prompts, and your Jina API key to Jina AI. Avoid using it for secrets, internal-only URLs, private documents, or authenticated content unless you intend that data to leave your environment.
Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.
)
try:
with urllib.request.urlopen(req, timeout=120) as resp:
content = resp.read().decode("utf-8")
print(content)
except urllib.error.HTTPError as e:
The code’s behavior is narrowly focused on Jina DeepSearch and is consistent with the DeepSearch portion of the description. However, the declared purpose presents the skill more broadly as covering web reading, web search, and deep research. In the supplied code chunk, only the deep research capability is implemented. Since the evaluation asks whether the declared description accurately represents what the supplied code chunk actually does, this is a description-behavior mismatch due to materially broader declared capabilities than the code demonstrates.
The code chunk is narrowly focused on one capability: searching via s.jina.ai. While that is part of the declared description, the overall declared purpose claims additional capabilities—reading webpages through r.jina.ai and performing deep multi-step research—that are absent from this code. There are no extra undeclared sensitive behaviors, and the accessed resource (s.jina.ai) is consistent with one declared function, but the description does not accurately represent this specific code chunk because it overstates the implemented functionality.
Referenced artifact was not completely inspected
Or use the helper script: `scripts/jina-search.sh "<query>" [--json]`
Referenced artifact was not completely inspected
Or use the helper script: `scripts/jina-search.sh "<query>" [--json]`
The README explicitly promotes URL fetching, web search, and deep research against Jina AI endpoints but does not warn users that supplied URLs, search queries, and potentially sensitive research prompts will be transmitted to third-party services. In a skill ecosystem, this omission can cause users or higher-level agents to unknowingly send confidential data, internal URLs, or sensitive queries off-system, creating privacy and data-handling risk.
The skill declares capabilities implying environment access, network access, and shell execution, but it does not define an explicit tool scope such as permissions or allowed-tools. That weakens least-privilege controls and makes autonomous invocation riskier, because the model may be able to use broader capabilities than a user expects from the manifest alone.
The skill sends user-provided URLs, queries, and an API key to external Jina-controlled endpoints. Although this is core to the skill's purpose and is openly disclosed, it still creates a real data-exfiltration surface if users pass sensitive URLs, authenticated targets, research prompts, or cookies via supported headers such as X-Set-Cookie.
# Site-scoped search
curl -s "https://s.jina.ai/OpenAI+GPT-5?site=reddit.com" \
-H "Authorization: Bearer $JINA_API_KEY" \
-H "Accept: text/plain"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
# Escape the question for JSON
JSON_QUESTION=$(printf '%s' "$QUESTION" | python3 -c 'import sys,json; print(json.dumps(sys.stdin.read()))')
response=$(curl -s -w "\n%{http_code}" "https://deepsearch.jina.ai/v1/chat/completions" \
-H "Authorization: Bearer $JINA_API_KEY" \
-H "Content-Type: application/json" \
-d "{
No suspicious patterns detected.