The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines and uses a shell command vexor for intent-based file discovery, incorporating user-provided queries and directory paths into the execution.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes local file contents to generate search results for the agent. Ingestion points: Local repository files via the vexor search results. Boundary markers: Absent. Capability inventory: Shell command execution and file discovery. Sanitization: Absent.
[DATA_EXFILTRATION]: The skill describes flags such as --include-hidden and --no-respect-gitignore for the vexor tool, which enable the agent to access sensitive files typically excluded from standard searches, such as .env files or SSH keys. Furthermore, the troubleshooting instructions mention connectivity and API configuration, indicating that the tool transmits data to a remote service to perform its semantic indexing and search functions.

vexor-cli