The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download the Coder CLI installation script from https://coder.com/install.sh, which is the official domain for the workspace platform.
[REMOTE_CODE_EXECUTION]: The installation workflow involves piping the downloaded script directly into a shell execution environment (curl ... | sh). This represents a standard setup procedure for the vendor's tooling.
[COMMAND_EXECUTION]: The skill's core functionality relies on executing coder CLI commands and ssh commands to manage lifecycle events (create, start, stop, delete) and perform development tasks within remote workspaces.
[PERSISTENCE]: The coder config-ssh -y command modifies the local ~/.ssh/config file to insert host entries. This modification persists across user sessions to enable seamless access to workspaces for various development tools.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from external systems that could be manipulated to influence the agent's behavior.
Ingestion points: The agent reads output from coder list, coder show, coder logs, and the results of commands executed via coder ssh (SKILL.md).
Boundary markers: There are no specific delimiters or instructions provided to differentiate between intended command output and potentially malicious embedded instructions in the ingested data.
Capability inventory: The skill allows for arbitrary command execution on remote workspaces, file reading/writing, and modification of local SSH configuration (SKILL.md).
Sanitization: The instructions do not define any sanitization or validation logic for the content returned from remote commands or workspace logs.

workspaces