The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the Datex Studio CLI (e.g., dxs source explore configs --branch <branchId>). The use of the <branchId> variable directly in the command string creates a command injection vulnerability if the input provided to the agent is not sanitized and contains shell metacharacters.
[DATA_EXFILTRATION]: The skill performs a bulk export of project configuration files to a temporary directory ($TMP/dxs-validate) using dxs source document build. While this is part of the intended linting workflow, it involves reading and duplicating sensitive project metadata and code structure to a temporary location on the local file system.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it parses and analyzes untrusted data from the codebase (JSON/YAML configurations and embedded TypeScript code).
Ingestion points: Files located in $TMP/dxs-validate/ exported from the branch.
Boundary markers: None identified; the agent reads the file content directly for heuristic analysis.
Capability inventory: Shell command execution via dxs CLI and file system access.
Sanitization: No explicit sanitization or validation of the ingested configuration content is described before analysis.

project-validator