The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted natural language descriptions to generate executable code. Ingestion points: User input is accepted through the --input CLI flag and the task field in API requests (documented in SKILL.md). Boundary markers: No clear delimiters or instructions to ignore embedded commands are present in the skill's usage patterns. Capability inventory: Documentation outlines capabilities for writing files and executing shell-based CSS processors. Sanitization: There is no mention of input validation or sanitization to prevent adversarial instructions from influencing the output.
[COMMAND_EXECUTION]: The integration notes provide examples of shell command execution, specifically suggesting that users pipe output to tools like postcss or use openclaw-sdk to run tasks. This encourages a workflow where the agent interacts with the underlying operating system based on generated content.

web-css