plugdata-patch
Warn
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill provides instructions and API references for creating custom audio objects using the
pd-luaextension, as detailed inreferences/pd-lua.md. The agent is guided to generate and save.pd_luafiles containing logic that is executed by the host's Lua interpreter during patch operation.- [COMMAND_EXECUTION]: The reference filereferences/hvcc.mddescribes the use of the Heavy compiler (hvcc) and includes specific command-line examples such ashvcc _main.pd -o output/ -n ProjectName -g c dpf. This invites the agent to perform shell-level operations to compile patches into executable plugin formats or embedded firmware targets.- [INDIRECT_PROMPT_INJECTION]: The skill's primary function of generating code from user descriptions creates a surface for indirect prompt injection. - Ingestion points: User-provided strings describing synthesizer architecture or effect parameters are interpolated directly into functional
.pdpatches and Lua scripts. - Boundary markers: There are no instructions for using delimiters or explicit warnings to the agent to treat user data as untrusted content.
- Capability inventory: The agent has the ability to write arbitrary files to the disk and execute local shell commands for compilation.
- Sanitization: The provided documentation lacks any mention of sanitizing, escaping, or validating user-provided input before it is used to generate logic or command-line arguments.
Audit Metadata