Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward Twitter/X search helper that uses the user's bearer token to query X's API and return tweet JSON, with no hidden code or persistence beyond token configuration guidance.
Install only if you are comfortable providing a Twitter/X API bearer token and paying X API usage costs. Avoid using confidential names, incidents, customer data, or secrets as search keywords, especially from automated cron or agent workflows.
The trigger language is very broad, including generic phrases and automatic invocation by cron jobs or other agents, which can cause the skill to run in contexts the user did not clearly intend. Because this skill sends search terms to an external API and incurs API cost, unintended invocation can lead to privacy leakage of user-provided topics and unnecessary spend.
The skill description does not clearly warn that caller-supplied keywords will be transmitted to X/Twitter's external API. If users or upstream agents pass sensitive names, incident details, or internal project terms, that data may be disclosed to a third party without informed consent.
The quick-reference example demonstrates immediate transmission of the query and token-authenticated request to X/Twitter's API. While this is core functionality rather than malicious behavior, it confirms the skill sends data off-platform and therefore creates third-party disclosure and cost-incurring risk if invoked carelessly.
curl -s -G "https://api.x.com/2/tweets/search/recent" \
--data-urlencode 'query=("AI agent" OR "agentic AI") (lang:en) -is:reply -is:retweet' \
--data-urlencode "max_results=30" \
--data-urlencode "start_time=$(date -u -v-24H +%Y-%m-%dT%H:%M:%SZ)" \
The parameters table sets the default 'languages' value to '["en"]', which means the skill will return English-only results unless the caller overrides it. This is a natural-language policy concern because it imposes a locale preference without explicit user opt-in or justification.
This second E1 finding at the same location reflects the same underlying behavior: outbound transmission to a third-party API. In the context of a search skill, the danger is not the network call itself but the absence of safeguards around what data may be included in the transmitted query.
Use start_time as an API parameter for the time window:
curl -s -G "https://api.x.com/2/tweets/search/recent" \
--data-urlencode "query=QUERY_HERE" \
--data-urlencode "max_results=50" \
--data-urlencode "start_time=ISO_TIMESTAMP" \
This second E1 finding at the same location reflects the same underlying behavior: outbound transmission to a third-party API. In the context of a search skill, the danger is not the network call itself but the absence of safeguards around what data may be included in the transmitted query.
Use start_time as an API parameter for the time window:
curl -s -G "https://api.x.com/2/tweets/search/recent" \
--data-urlencode "query=QUERY_HERE" \
--data-urlencode "max_results=50" \
--data-urlencode "start_time=ISO_TIMESTAMP" \
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
GET https://api.x.com/2/tweets/search/recent
Searches tweets from the last 7 days.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
GET https://api.x.com/2/tweets/search/recent
Searches tweets from the last 7 days.
The quick reference hard-codes 'lang:en', which implies English-only results by default. Forcing a specific language without explicit user choice can violate language/locale policy expectations unless the restriction is documented as optional or justified.
No suspicious patterns detected.