The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill retrieves a Telegram bot token by reading a local configuration file at ~/.openclaw/openclaw.json within the scripts/send_telegram.sh script.
[COMMAND_EXECUTION]: The skill executes multiple local scripts (screenshot.js, send_telegram.sh) and uses node -e for inline JavaScript execution. It also dynamically searches for and requires the playwright module from the user's ~/.npm/_npx directory, which constitutes dynamic module loading from a user-writable path.
[EXTERNAL_DOWNLOADS]: The skill requires downloading and installing the Playwright library and Chromium browser, which are external dependencies hosted on public registries.
[DATA_EXFILTRATION]: The scripts/send_telegram.sh script transmits the generated infographic and user-provided captions to the Telegram API (api.telegram.org). While this is a documented feature, it represents an automated transfer of local data to an external service using local credentials.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted external content and embeds it into HTML without explicit sanitization.
Ingestion points: Articles and tweets processed by the agent in SKILL.md.
Boundary markers: Absent; the content is placed directly into the HTML structure.
Capability inventory: File system access, script execution, and network communication via the included Node.js and Bash scripts.
Sanitization: No evidence of content sanitization or escaping is provided in the skill instructions, potentially allowing the execution of scripts within the rendering environment.

infographic