hackernews

Fail

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • CREDENTIALS_UNSAFE (HIGH): The skill reads raw usernames and passwords from a local file (community-pulse/skills/hackernews/.env.profiles.local). Storing and accessing unencrypted secrets on the filesystem is a high-severity security risk.
  • PROMPT_INJECTION (LOW): The skill is susceptible to Indirect Prompt Injection because it processes untrusted content from Hacker News to draft responses. * Ingestion points: Thread data (titles and comments) is ingested in research.md using browser tools. * Boundary markers: Absent; there are no delimiters to separate untrusted data from the agent's instructions. * Capability inventory: The skill uses Browser MCP tools (navigate, form_input, computer) to perform actions based on the processed data. * Sanitization: No sanitization or validation of external web content is mentioned in the skill definition.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 19, 2026, 01:55 AM