The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The install.sh script downloads and installs multiple CLI tools (clj-nrepl-eval, clj-paren-repair-claude-hook, clj-paren-repair) using bbin from an external GitHub repository (https://github.com/bhauman/clojure-mcp-light.git) which is not an identified trusted vendor.
[REMOTE_CODE_EXECUTION]: The skill is designed to evaluate arbitrary Clojure code on a running system via clj-nrepl-eval. This enables the execution of commands and logic within the context of the application being debugged or managed.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands to discover local network ports and interact with the nREPL server.
[DYNAMIC_EXECUTION]: The core functionality relies on the runtime evaluation of Clojure forms (eval) to inspect namespaces, vars, and live application state.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from the REPL which could contain instructions if the target environment contains untrusted data.
Ingestion points: Response data from clj-nrepl-eval commands in SKILL.md.
Boundary markers: None identified for delimiting REPL output from agent instructions.
Capability inventory: Execution of arbitrary Clojure code, local shell command execution, and discovery of local network services.
Sanitization: No sanitization or filtering of REPL output is mentioned before processing by the agent.

clojure-repl