The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/tts_expressive.py dynamically loads and executes a Python module from a path defined by the DOUBAO_VOICE_PY environment variable using importlib.util.module_from_spec. This allows the execution of arbitrary Python code if the environment variable is manipulated to point to a malicious file.
[COMMAND_EXECUTION]: The skill extensively uses subprocess.run to execute external binaries including curl, ffmpeg, and ffprobe. Additionally, SKILL.md instructs the agent to run various shell commands (npm, node, bash) with parameters derived from user-provided titles or external data, which creates a potential surface for command injection if inputs are not strictly validated.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to search for and ingest content from the public web and media repositories to generate scripts. It lacks robust sanitization or boundary markers to prevent instructions embedded in that external content from influencing the agent's behavior.
Ingestion points: Web search results and content fetched via curl (specified in SKILL.md step 1 and scripts/commons.py).
Boundary markers: No delimiters or instructions to ignore embedded commands were found in the prompt construction.
Capability inventory: File system writes, network operations via curl, and execution of local scripts and system binaries (ffmpeg, node, npm).
Sanitization: The skill performs basic HTML tag removal for readability, but no security-focused sanitization of the text is implemented.
[EXTERNAL_DOWNLOADS]: The scripts/commons.py script downloads images and videos from Wikimedia Commons (commons.wikimedia.org). While this is a well-known service, the skill's ability to fetch and store arbitrary files based on external search results is a notable security surface.
[AUTONOMY_RISK]: The instructions in SKILL.md explicitly direct the agent to complete steps 0 through 8 "in one go" without waiting for user confirmation. This high level of autonomy reduces human oversight and could allow malicious operations or significant errors to proceed without intervention.

weilanx-creator-vox-video