The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command python3 seedance_lint.py "<prompt text>" using raw input from the user. This creates a critical command injection vulnerability where a user could provide a prompt containing shell metacharacters (e.g., ;, &, |) to execute arbitrary commands on the host system.
[PROMPT_INJECTION]: The skill's primary function is to teach the agent how to help users circumvent content filters for Seedance 2.0. It provides a 'Rewrite Playbook' and specific conversion rules (e.g., transforming 'violence' into 'aftermath' or 'physics') to evade safety guardrails.
[COMMAND_EXECUTION]: The instruction to use higgsfield_memory.py to append entries to a local database based on user-provided data indicates a capability for persistent file system modification which can be influenced by untrusted input.
[COMMAND_EXECUTION]: Analysis of the Indirect Prompt Injection surface (Category 8): (1) Ingestion points: User-supplied prompt text (SKILL.md); (2) Boundary markers: Absent; (3) Capability inventory: Subprocess execution via python3; (4) Sanitization: Absent.

higgsfield-seedance