The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The README and install.sh files promote a high-risk installation pattern where a remote shell script is downloaded and piped directly into bash (curl -fsSL https://raw.githubusercontent.com/0xNyk/xint/main/install.sh | bash).
The script originates from a personal GitHub repository (0xNyk) which is not identified as a trusted organization or well-known service.
The installation process subsequently triggers unpinned dependency installation via bun install at runtime.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted tweet data from X and processing it via the Grok AI model for sentiment analysis and reporting.
Ingestion points: lib/api.ts (tweet search and profile results).
Boundary markers: Present in lib/sentiment.ts using a bracketed ID and username format ([id] @username: text).
Capability inventory: The skill possesses network read capabilities and subprocess execution via Bun.spawn.
Sanitization: Basic ANSI escape sequence removal is performed in lib/tui.ts before display.
[COMMAND_EXECUTION]: The TUI (Terminal User Interface) implementation in lib/tui.ts utilizes Bun.spawn to execute CLI subcommands.
These commands are dynamically constructed in lib/tui_adapter.ts based on user input and history categories.
[SAFE]: Sensitive credentials such as X_BEARER_TOKEN and XAI_API_KEY are managed through environment variables or local .env files. OAuth tokens are stored in data/oauth-tokens.json with restrictive permissions (chmod 600). Webhook functionality includes protocol validation in lib/webhook-security.ts to ensure remote endpoints use HTTPS.

xint