The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell curl to fetch content from user-provided URLs (e.g., blog posts, gists). Using shell commands with unsanitized user input poses a high risk of command injection if the URL contains shell metacharacters.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest data from external, untrusted sources and use it as verbatim prompts for video and image generation. This creates an attack surface where malicious text in the fetched document could control the agent's generative outputs.
Ingestion points: External URLs fetched using curl in Step 0 (SKILL.md).
Boundary markers: Absent; the skill explicitly forbids modifying or 'cleaning' the input.
Capability inventory: Includes shell command execution (curl), image generation, video generation, and project rendering (SKILL.md).
Sanitization: None; the instructions require 'word-for-word' execution.
[EXTERNAL_DOWNLOADS]: The skill fetches content from arbitrary external domains provided by the user, which can be used to pull malicious content into the agent's context.

flowy-verbatim