The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download and install a compiled binary from an external GitHub repository (github.com/mvanhorn/cli-printing-press) using the go install command.
[COMMAND_EXECUTION]: The installation process involves running shell commands (go install, git clone, git pull) to fetch and execute code from sources outside the vendor's own infrastructure.
[DYNAMIC_EXECUTION]: The skill is designed to autonomously generate Go code, compile it, and then execute it during its "Verification" and "Dogfood" phases. This process involves running exec.Command on dynamically created binaries.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data—such as URLs, third-party API documentation, and "browser-sniffed" network traffic—to drive its autonomous code generation workflow.
Ingestion points: Inputs to the /printing-press <url> command and automated web research phases (File: SKILL.md).
Boundary markers: The instructions lack explicit directives for the agent to ignore or sanitize malicious instructions that might be embedded within targeted API specifications or captured web traffic.
Capability inventory: The skill has the capability to write files, compile Go code, and execute subprocesses (exec.Command) to test the generated output (File: SKILL.md).
Sanitization: There are no mentioned mechanisms for sanitizing or escaping content retrieved from external URLs before it is interpolated into the code generation pipeline.

cli-printing-press-generator