The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill directs the agent to download and immediately execute a script from a remote URL using the curl -fsSL https://jimeng.jianying.com/cli | bash command. This pattern is a high-severity risk as it allows for unverified code to be executed with the privileges of the shell. This finding contributes to a HIGH verdict because it is a primary installation method for the skill's required tooling.
[EXTERNAL_DOWNLOADS]: The skill fetches a CLI tool from a third-party domain (jimeng.jianying.com). While the domain belongs to a known service, the unverified download and execution method bypasses standard package security.
[PROMPT_INJECTION]: The skill uses WebFetch to gather market intelligence and industry articles from external websites (Phase 0.2). This ingestion of untrusted data lacks boundary markers or sanitization, creating an attack surface for indirect prompt injection.
Ingestion points: Industry rankings and analysis articles fetched via WebFetch in Phase 0.2.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the data processing workflow.
Capability inventory: Shell command execution, Python subprocess calls, and CLI tool interaction.
Sanitization: No validation or filtering is applied to the content retrieved from external URLs before it enters the agent's context.
[COMMAND_EXECUTION]: The skill extensively uses shell commands and Python scripts (subprocess) to manage project files, verify tool status, and generate creative assets.

ai-short-drama