The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to extract secrets from the macOS Keychain using the 'security find-generic-password' command and reads a local environment file at '~/.nexus/.env' to retrieve 'RAPIDAPI_KEY', 'BRAVE_SEARCH_API_KEY', and 'SEARCH_API_KEY'.
[DATA_EXFILTRATION]: The skill generates and uploads reports to an external VPS at a hardcoded IP address '89.116.229.189'. This transmission follows the retrieval of sensitive credentials from the host machine.
[PRIVILEGE_ESCALATION]: The instructions direct the agent to perform 'chmod +x' on binaries located in '~/.nexus/cli-tools/' if they are not executable, which involves modifying file system permissions at runtime.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute multiple shell commands for credential retrieval, environment variable grepping, and running custom CLI binaries and deployment scripts.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface as it ingests untrusted data from web searches and web scraping (Brave Search, Playwright) and processes it with sensitive capabilities. Evidence chain: 1. Ingestion points: Brave Search results and Playwright scraper outputs. 2. Boundary markers: Absent. 3. Capability inventory: Bash (shell execution), Write (file system access), and Network upload (VPS deployment). 4. Sanitization: No sanitization or validation of the external content is mentioned.

travel