The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted code from local directories and remote Git repositories, creating a surface for potential prompt injection via code comments or documentation.
Ingestion points: Local files passed via the -f flag and remote repositories cloned during the PR review workflow as described in SKILL.md.
Boundary markers: The instructions and examples in SKILL.md do not specify delimiters or explicit warnings for the agent to ignore instructions embedded in the code it processes.
Capability inventory: The skill leverages the terminal tool for shell commands and the process tool for handling interactive TUI sessions in SKILL.md.
Sanitization: There are no explicit sanitization or validation steps mentioned within the skill for content retrieved from external sources before it is passed to the OpenCode tool.
[EXTERNAL_DOWNLOADS]: The skill describes how to install and use the OpenCode tool from standard registries and repositories.
Suggests installation of the opencode-ai package via npm and the opencode formula via Homebrew (SKILL.md).
Provides examples of cloning repositories from GitHub for code review purposes (SKILL.md).

opencode