The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to install the opencode-ai package using global installation commands (npm i -g opencode-ai@latest or brew install anomalyco/tap/opencode). Using unpinned versions (e.g., @latest) introduces a potential supply chain risk where a compromised package update could affect the environment.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core functionality of processing external codebases.
Ingestion points: Untrusted data enters the context when the skill is used to review pull requests (opencode pr), clone remote repositories, or read project files containing external content.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between its own system instructions and potentially malicious instructions embedded in the code files being analyzed.
Capability inventory: The skill possesses extensive capabilities through the terminal and process tools, allowing it to execute shell commands and modify the local filesystem.
Sanitization: The instructions do not specify any sanitization or validation steps for the content ingested from external repositories.
[COMMAND_EXECUTION]: The skill makes heavy use of terminal commands to interact with the OpenCode CLI, manage git repositories, and handle background processes. While these are required for its primary purpose, the broad use of terminal(command=...) with variable input (like PR numbers or project paths) necessitates strict user oversight.
[DYNAMIC_EXECUTION]: By design, this skill acts as an autonomous coding worker that generates and executes code to refactor, debug, or implement features. This involves runtime generation and execution of scripts within the working directory.

opencode