finding-data-scientists-on-twitter
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data from Twitter (bios, tweets, and profiles). This content is used to score candidates and generate recruitment lists, creating a surface where malicious instructions embedded in social media profiles could potentially influence the agent's behavior.
- Ingestion points: Twitter profiles and tweet content fetched via the Apify API (SKILL.md).
- Boundary markers: None identified; the skill directly processes bio and tweet text for keyword matching and scoring.
- Capability inventory: The skill uses
curland local Node.js scripts to interact with the Apify API and generates candidate reports (SKILL.md). - Sanitization: No explicit sanitization or filtering of the ingested Twitter content is mentioned before it is processed for scoring.
- [DYNAMIC_EXECUTION]: The skill exposes a
customMapFunctioninput parameter, which accepts a JavaScript function as a string. This function is intended to transform output objects during the scraping process on the Apify platform, representing a mechanism for dynamic code execution. - [COMMAND_EXECUTION]: The documentation references and provides examples for executing a local script
scripts/run_actor.jsusing Node.js. This script acts as a wrapper for interacting with the Apify API, handling actor execution, and saving output to local files (CSV/JSON).
Audit Metadata