effectcraft-lower-thirds
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides instructions and Python code to download an external binary executable (
effectcraft-cli.exe) from a third-party GitHub repository (storytold/effectcraft) and run it on the local system. This creates a pathway for executing unverified code. - [EXTERNAL_DOWNLOADS]: A direct link to a portable ZIP archive containing a binary executable from a non-whitelisted source is provided in
references/craft-command-evidence.json. - [COMMAND_EXECUTION]: The skill includes a Python automation script in
references/animated-lower-third.mdthat utilizessubprocess.runto execute external shell commands with parameters. - [DYNAMIC_EXECUTION]: The provided Python script dynamically builds command-line arguments by reading and parsing a local JSON file (
references/animated-lower-third.json) at runtime, which is then passed to a shell execution tool. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection. Ingestion points: The skill reads external video assets and user-provided metadata (names and roles) as described in
SKILL.md. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present. Capability inventory: The skill has access tosubprocess.runand tool execution viacommand_run. Sanitization: No evidence of input validation or escaping for user-provided text before it is used in command parameters.
Audit Metadata