The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill explicitly suggests executing code from https://opencode.dev/install by piping it directly to bash (curl -fsSL ... | bash). This is a high-risk pattern as it allows for arbitrary code execution from an untrusted external source without verification.
[COMMAND_EXECUTION]: The skill relies on the bash tool to interact with the local environment, including running the opencode CLI and managing process lifecycles, which poses a risk if these commands are manipulated by external input.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the output of the OpenCode process which could be influenced by external project data or attacker-controlled prompts.
Ingestion points: Data is ingested via process.poll in SKILL.md.
Boundary markers: There are no boundary markers or instructions to the agent to ignore embedded commands within the tool output.
Capability inventory: The skill has access to bash, process.write, and process.kill as seen in SKILL.md.
Sanitization: There is no evidence of output filtering or sanitization before the data is processed by the agent.

opencode-acp-control