x-tweet-search
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PERSISTENCE]: The
scripts/setup.shscript provides instructions that modify the user's shell profile (~/.zshrc) to store theX_BEARER_TOKEN. While this is for legitimate configuration, it utilizes a persistence mechanism to maintain environment variables across sessions. - [INDIRECT_PROMPT_INJECTION]: The skill fetches raw content from X/Twitter, an untrusted external source. The retrieved data is returned to the agent without sanitization or boundary markers, which could allow malicious instructions embedded in tweets to influence the agent's behavior.
- Ingestion points: Fetches data from
https://api.x.com/2/tweets/search/recentinscripts/search.sh. - Boundary markers: Absent; the script returns raw JSON output directly to the agent context.
- Capability inventory: The skill uses
curlandjqvia Bash, with the ability to perform network requests. - Sanitization: None; the tweet content is passed through without filtering.
Audit Metadata