The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes a helper script scripts/openstory.sh that checks for the existence of specific environment variable names in the .env.local file.\n
Evidence: The script uses grep -qE to detect keys such as FAL_KEY, OPENROUTER_KEY, and BETTER_AUTH_SECRET.\n
Analysis: The script is designed to be read-only and only reports whether the keys are "set" or "unset" without outputting the secret values or exfiltrating data.\n- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The instructions guide the user to clone the OpenStory repository from GitHub and install dependencies.\n
Evidence: git clone https://github.com/openstory-so/openstory.git and bun install commands.\n
Analysis: These are standard setup procedures for the open-source project that the skill is designed to support. GitHub is a well-known and trusted service.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process user-supplied scripts into video sequences, which presents a surface for indirect prompt injection.\n
Ingestion points: The StoryboardWorkflow processes scripts and user-provided inputs to generate visual and motion prompts.\n
Capability inventory: The skill utilizes Bash execution and file modification tools.\n
Boundary markers: Not explicitly defined in the provided instructions.\n
Sanitization: Not explicitly documented in the skill's logic.\n
Analysis: This is an inherent risk for agents processing untrusted data, though the skill itself does not contain malicious triggers.

openstory