social-scraping
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Requires the agentcash package from the NPM registry to perform authenticated API requests and handle payments.
- [COMMAND_EXECUTION]: Executes shell commands via npx agentcash@latest to trigger data collection and poll for results from the StableSocial service.
- [PROMPT_INJECTION]: Risk of indirect prompt injection through ingested social media data. 1. Ingestion points: Retrieves posts, comments, and profile data from TikTok, Instagram, Facebook, Reddit, and LinkedIn via endpoints listed in SKILL.md. 2. Boundary markers: Absent; there are no instructions to the agent to treat scraped content as untrusted or to ignore instructions within it. 3. Capability inventory: The skill utilizes network access and command execution through the agentcash CLI as documented in SKILL.md and rules/getting-started.md. 4. Sanitization: Absent; no sanitization or filtering of the retrieved social media content is performed before presentation to the agent.
Audit Metadata