The Agent Skills Directory

[PERSISTENCE]: The installation script (bin/cli.js) includes functionality to modify shell initialization files, such as .bashrc, .zshrc, or config.fish, to store the EVOLINK_API_KEY environment variable across sessions. The script implements a user confirmation step before making these modifications.
[COMMAND_EXECUTION]: The skill's installation and operational logic involve the execution of shell commands. The installer (bin/cli.js) uses child_process.execSync and spawnSync for environment detection and fs.chmodSync to set executable permissions on scripts. The primary functionality involves the agent executing the scripts/seedance-gen.sh bash script.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, creating a potential surface for indirect prompt injection.
Ingestion points: The skill ingests user-provided text prompts and external URLs for images, videos, and audio files via command-line arguments in SKILL.md and scripts/seedance-gen.sh.
Boundary markers: The skill does not employ specific delimiters or "ignore instructions" warnings to isolate user-provided data from system instructions.
Capability inventory: The skill can execute local shell scripts and perform network requests via curl to the api.evolink.ai endpoint.
Sanitization: The generation script utilizes jq with argument binding (--arg and --argjson) to safely construct JSON payloads, which prevents basic JSON injection but does not mitigate malicious content within the prompt or media metadata from influencing the underlying AI model.

seedance-2-video-gen