The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md direct the agent to install a tool using a high-risk remote execution pattern. Specifically, it uses curl -fsSL https://vite.plus | bash. This command downloads a script from an external, unknown domain and executes it immediately in the shell without any verification, integrity checks, or version pinning. This represents a significant risk as the remote content can be modified to execute arbitrary malicious code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze project evidence and user preferences to generate executable configurations and run toolchain verification commands, creating a vulnerability surface.
Ingestion points: The skill reads project-specific evidence and directory structures from ~/workspace to inform its technical recommendations and configuration generation.
Boundary markers: There are no explicit delimiters or instructions to prevent the agent from obeying malicious instructions embedded within the analyzed project "evidence."
Capability inventory: The skill possesses the capability to modify pyproject.toml files and execute a variety of CLI tools including uv, ruff, ty, pytest, and the untrusted vp tool.
Sanitization: No sanitization or validation of the input evidence is performed before it is used to influence the generation of local commands and configuration files.

tech-preferences