The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data sources which could potentially contain malicious instructions intended to influence the agent's behavior.
Ingestion points: Data is ingested from the Undercroft platform (tenants, runs, dbt models, reports) and from GitHub issue lists when checking for existing bug reports.
Boundary markers: The skill does not implement specific delimiters or "ignore" instructions for the platform data it retrieves.
Capability inventory: The skill utilizes shell command execution (CLI tools), network access for API calls, and file system interactions for configuration.
Sanitization: The skill includes a mandatory redaction step in the bug reporting process, instructing the agent to replace real IDs and credentials with placeholders like CASE-0042 and acme@example.test before posting to public repositories.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of a CLI tool from a remote repository.
Finding: Downloads the undercroft-cli package from the author's GitHub repository.
Evidence: https://github.com/muitneliss/undercroft/releases/download/v1.22.1/undercroft-cli-1.22.1.tgz
[COMMAND_EXECUTION]: The skill instructs the agent to execute several system commands to perform its tasks.
Tools: npm, node, undercroft, gh, and curl.
Context: These tools are used for package installation, version validation, platform management, and automated issue creation on GitHub.

undercroft-cli