finding-marketing-professionals-on-twitter

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The customMapFunction input parameter accepts a string intended to be a JavaScript function. This allows for the execution of arbitrary code within the context of the scraping actor's environment to transform output data.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external sources (Twitter bios and tweets) to generate candidate scores and summaries.
  • Ingestion points: Bio metadata and tweet text retrieved from the apidojo~twitter-user-scraper and apidojo~tweet-scraper actors in SKILL.md.
  • Boundary markers: Absent. There are no explicit instructions to the agent to treat scraped bio content as untrusted data or to ignore embedded instructions.
  • Capability inventory: The skill uses scraped data to perform scoring logic and generate candidate lists in the chat output.
  • Sanitization: None described. The agent is instructed to use free-text bios directly for keyword matching and scoring.
  • [COMMAND_EXECUTION]: The skill instructions include multiple examples of shell commands involving node scripts/run_actor.js. This encourages the agent to execute a local script and provides a template for passing arguments to a shell environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 07:05 PM
Security Audit — agent-trust-hub — finding-marketing-professionals-on-twitter