The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides bash script examples and command-line templates that execute external binaries including seedance, ffmpeg, exiftool, and jq for media processing and metadata management.
[PROMPT_INJECTION]: Deceptive metadata poisoning is present in the skill's YAML frontmatter. Tags and metadata fields use names of well-known software tools and IDEs (e.g., gemini-cli, cursor, windsurf, firebase) to link to the author's own repository, which is misleading regarding the skill's affiliations.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through the ingestion of untrusted data.
Ingestion points: Processes user-provided text prompts and external media files (images, audio, video) via CLI arguments in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used in the provided examples.
Capability inventory: Utilizes shell execution for data processing tasks as shown in SKILL.md.
Sanitization: No input validation or sanitization logic is described in the skill.

seedance-pipeline