The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from $ARGUMENTS or previous conversation history and interpolates it directly into a JavaScript context within a local HTML file. * Ingestion points: $ARGUMENTS and conversation history (SKILL.md). * Boundary markers: Absent. * Capability inventory: Uses the 'Write' tool to modify ~/.claude/skills/speed/data/reader.html and the 'Bash' tool to execute the open command. * Sanitization: Relies on the agent to manually 'Escape quotes and backslashes', which is an unreliable security control against adversarial input designed to break out of JavaScript string literals.
[COMMAND_EXECUTION]: The skill uses the 'Bash' tool to run the open command on a local file path. This action triggers the execution of the dynamically generated HTML/JavaScript content in the user's default web browser.

speed