The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends global installation of the opencode-ai package via npm (npm install -g opencode-ai@latest). This introduces a dependency on an external package for its core functionality.\n- [COMMAND_EXECUTION]: The skill invokes the opencode binary and uses local Python scripts (delegation_executor.py) to process tasks. It also references a Makefile to orchestrate delegation calls.\n- [INDIRECT_PROMPT_INJECTION]: The skill creates a surface where external input is passed to a terminal-capable agent.\n
Ingestion points: User-provided strings (PROMPT) and file contents are inlined into the opencode run command as described in SKILL.md.\n
Boundary markers: Absent; the prompt is passed as raw positional arguments to the CLI tool.\n
Capability inventory: The opencode agent is explicitly described as having code_execution capabilities, posing a risk if it executes malicious code embedded in a prompt.\n
Sanitization: Absent; inputs are directly interpolated into the command string without validation or escaping.

opencode-delegation