The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted content from the context argument (which can be a URL, file path, or raw text) to generate narration scripts and image prompts.
Ingestion points: The $1 (context) argument defined in SKILL.md is used as the primary source of truth for video content.
Boundary markers: There are no explicit delimiters or boundary instructions provided to the agent to isolate the context data from the skill's instructions.
Capability inventory: The skill has access to the Bash tool for running project commands and the Write tool for creating the story.json configuration.
Sanitization: The workflow includes a mandatory human review step (Step 5) where the narration must be approved before generation proceeds, though a "fully automatic alternative" exists that skips this safeguard.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local npm and npx commands that incorporate parameters derived from user input, such as the slug and CompositionId.
Evidence: Commands like npm run generate -- --slug <slug> and npx remotion render <CompositionId> ... directly interpolate user-supplied strings into shell executions, which could lead to command injection if input is not strictly validated by the agent.

story-new