The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/setup/filmcraft-update script is designed to fetch release assets and binaries from the storytold/filmcraft and storytold/effectcraft repositories on GitHub. These repositories are not among the trusted vendor or organization lists provided in the security policy.
[REMOTE_CODE_EXECUTION]: The skill facilitates the download and execution of binary software from unverified external sources via the gh release download command and subsequent installation to the local system path.
[PRIVILEGE_ESCALATION]: The scripts/setup/filmcraft-update utility utilizes pkexec or sudo -A to acquire root privileges for the purpose of installing software to protected system directories such as /opt and /usr/local/bin.
[COMMAND_EXECUTION]: A significant number of scripts (e.g., build_hook1.py, talk_motion.py, add_music.py, fill_jumpcuts.py) dynamically construct and execute shell commands using subprocess.run and subprocess.Popen. These commands interact with filmcraft-cli, ffmpeg, and ffprobe to manipulate video assets and project files. While intended for automation, this pattern represents a broad attack surface for command injection if input parameters are not strictly validated.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading and processing external data, including user-provided script Markdown files and tool-generated ASR transcripts (e.g., talking-head-enhanced.words.json). If these files contain malicious instructions, they could potentially influence the AI agent's generation of FilmCraft commands or editing logic, although the immediate impact is mitigated by standard LLM guardrails.

edit-style