The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a protocol for discovering and loading additional 'project skills' from various hidden folders in the workspace (such as .agents/skills/, .cursor/skills/, etc.) and user-level directories. It instructs the agent to search for, open, and follow the instructions in any SKILL.md files found. This creates an attack surface where hidden files in a repository could be used to inject malicious instructions into the agent's context.
Ingestion points: references/monorepo-contract.md (Section: Discovering project skills).
Boundary markers: No explicit markers or instructions are provided to separate or ignore instructions within the discovered skill files.
Capability inventory: The skill utilizes npx, pnpm, and turbo for shell command execution.
Sanitization: No validation or sanitization is performed on the content of the discovered skills.
[COMMAND_EXECUTION]: The skill instructs the agent to perform management tasks using shell commands such as pnpm, turbo, and npx. These tools provide the necessary capabilities for potential exploitation if the agent is influenced by instructions from the indirect prompt injection surface.
[EXTERNAL_DOWNLOADS]: The setup process involves downloading and installing additional skills from an external repository (antfu/skills) using the npx skills add command. This represents a dependency on external code to extend the agent's functionality.

managing-monorepo