The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a significant attack surface by ingesting untrusted data from various external sources including API documentation, web search results, and captured network traffic (HAR files) to build its internal model and generate code (documented in phases/03-resolve-and-reuse.md, phases/04-research-brief.md, and phases/08-ecosystem-absorb-gate.md).
Capabilities: The skill has extensive shell execution and file system access permissions.
Sanitization: It implements a detailed secret and PII redaction system in references/secret-protection.md to prevent accidental data exfiltration.
Boundary Markers: Uses explicit instructions and specialized tools like shipcheck to verify the safety and correctness of the generated output.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic code generation and execution as a core feature. It delegates Go code construction tasks to the Codex CLI using the --yolo flag for automated application of changes (references/codex-delegation.md). Additionally, it executes inline Swift scripts via subprocesses to interact with macOS frameworks (phases/11-build-the-goat.md) and runs arbitrary JavaScript within browser contexts during the API discovery phase (references/browser-sniff-capture.md).
[EXTERNAL_DOWNLOADS]: Fetches and executes setup and installation scripts from the author's GitHub repository (raw.githubusercontent.com/mvanhorn/cli-printing-press/main/scripts/install.sh) as part of the preflight environment verification (phases/01-preflight.md). It also utilizes npx to download and update agent skills globally (references/setup-checks.md).
[COMMAND_EXECUTION]: Performs extensive command-line operations to manage the Go development environment, including go build, go install, and custom binary execution for lifecycle management and testing (phases/01-preflight.md, phases/10-generate.md, phases/18-dogfood-testing.md).

printing-press