The Agent Skills Directory

[PERSISTENCE]: The script scripts/generation_api.py modifies shell startup profiles such as ~/.bashrc, ~/.zshrc, and ~/.profile. It appends or updates an export command for the GIGGLE_API_KEY to ensure the credential persists across terminal sessions.
Evidence: The setup_api_key function in scripts/generation_api.py explicitly writes to detected shell configuration paths.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-provided text prompts and external resource URLs (images, videos, and audio) for processing by an LLM-based video generation engine. This architecture is vulnerable to indirect prompt injection if malicious instructions are embedded in the fetched external data.
Ingestion points: The --prompt argument and media URLs provided through --images, --videos, and --audios in scripts/generation_api.py.
Capability inventory: The script performs network requests, writes to the home directory (~/.giggle_api_key), and modifies shell configuration files.
Sanitization: The skill does not appear to perform sanitization or validation of the content retrieved from external URLs before processing.
[COMMAND_EXECUTION]: The skill instructs the agent and user to execute a Python script (scripts/generation_api.py) to interact with the external API.
Evidence: SKILL.md contains multiple command-line examples for setting up the environment and generating video.
[DATA_EXFILTRATION]: The script reads a stored API key from ~/.giggle_api_key and transmits it via HTTP headers to https://giggle.pro for authentication.
Evidence: The SeedanceClient class in scripts/generation_api.py manages these network operations.
Context: The destination domain is a vendor-owned resource associated with the skill author, representing intended functionality.

giggle-seedance2-gen