The Agent Skills Directory

[Unverifiable Dependencies & Remote Code Execution] (LOW): The skill utilizes the clj-nrepl-eval tool to evaluate arbitrary Clojure code strings on a local nREPL server. \n
Evidence: Instructions in SKILL.md and references/tool-guide.md direct the agent to test and define functions via nREPL. \n
Context: Severity is lowered because code execution is the core intended functionality for a REPL-driven development skill. \n- [Indirect Prompt Injection] (LOW): The skill is susceptible to instructions embedded in project files that could influence agent behavior. \n
Ingestion points: The skill reads file content using the read command and searches through code with rg (ripgrep) as seen in SKILL.md. \n
Boundary markers: Absent; there are no instructions for the agent to ignore or isolate natural language instructions found in comments within Clojure or EDN files. \n
Capability inventory: Possesses the ability to write to the filesystem and execute arbitrary code via the REPL. \n
Sanitization: No validation or sanitization of file content is performed before the agent processes or executes it. \n- [Command Execution] (LOW): Uses system commands for code discovery and search. \n
Evidence: Documented use of rg and custom tools like clj-paren-repair in SKILL.md.

clojure-repl-dev