The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by instructing the agent to ingest and rely on external data from the codebase, such as package manifests, exports, and source code. An attacker could potentially embed malicious instructions within these files to influence the agent's behavior. The skill explicitly instructs the agent to treat this external source code and metadata as "authoritative," which may increase the likelihood of the agent obeying embedded adversarial instructions. Ingestion points: package.json files, source code files, and documentation in the references/ directory. Boundary markers: None specified; there are no instructions to use delimiters or ignore embedded instructions. Capability inventory: The agent is empowered to navigate the file system, read codebase files, and summarize architectural patterns. Sanitization: No sanitization or validation of the ingested codebase content is described.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for development workflows and codebase analysis. This includes project-specific scripts and search utilities. Evidence: The workflow in references/common.md recommends using ripgrep (rg) to search for exports and usage patterns (e.g., rg -n "@homarr/common" apps packages). SKILL.md and references/architecture.md mention application management commands such as bun run dev and bun run dev:cli.

codebase-context