The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted HTML, CSS, and JavaScript from user-provided files and live URLs. This creates a surface for potential instructions embedded in the source code to influence the agent's behavior.
Ingestion points: Analyzes local HTML files and external website content provided by the user.
Boundary markers: No specific boundary markers or instructions to ignore embedded directives are required for processing the source content.
Capability inventory: Possesses capabilities for file system writes, git operations, and network-enabled browser capture.
Sanitization: Input code is not sanitized or filtered before analysis.
[COMMAND_EXECUTION]: The skill uses local CLI tools for file verification and version control.
Git usage: Employs git for status checks, staging (including add -f for assets), and committing.
Media verification: Uses ffprobe to validate the readability of captured MP4 files.
Local serving: May launch a local server on localhost to facilitate browser access to local HTML files.
[DATA_EXFILTRATION]: The skill performs network operations to visit and capture content from user-provided URLs.
Network operations: Connects to external websites to record videos and take screenshots as part of its primary design documentation purpose.

html-to-interaction-prompts