The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from project files and user input which is then interpolated into prompts for sub-agents.
Ingestion points: Reads feature descriptions from user arguments and design documents from design/gdd/ and assets/audio/.
Boundary markers: The instructions lack explicit delimiters (e.g., XML tags or triple quotes) to separate untrusted file content from agent instructions.
Capability inventory: The skill and its sub-agents have access to Bash, Task (for agent spawning), and Write tools, providing a significant capability surface if malicious instructions were present in ingested files.
Sanitization: No evidence of escaping or validation for content retrieved from the filesystem before it is passed to sub-agents.
[COMMAND_EXECUTION]: The skill allows the use of the Bash tool and delegates integration tasks to sub-agents that may execute shell commands, though no specific dangerous commands are hardcoded in the instructions.

team-audio