The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads and installs binaries from github.com/mvanhorn/printing-press-library and the @mvanhorn/printing-press-library NPM package. These resources are maintained by the skill's vendor.
[DYNAMIC_EXECUTION]: The skill implements an 'automatic learning' loop where the agent is instructed to 'teach' the CLI new playbooks. These playbooks consist of stored shell command sequences (Playbook.steps) with variable substitution that are later replayed and executed by the agent during 'recall' operations. This creates a mechanism where previous inputs define future executable command sequences.
[DATA_EXFILTRATION]: The skill documents a --deliver webhook:<url> flag that allows the output of any CLI command to be transmitted to an arbitrary external URL via a POST request. This sink could be utilized to exfiltrate product data or information about the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Crestron.com during the sync process to build a local SQLite mirror. This external content is used to answer queries and guide agent actions, presenting a surface where malicious source data could influence agent behavior.
Ingestion points: Untrusted data enters via the crestron-pp-cli sync command, which crawls Crestron.com.
Boundary markers: The instructions do not specify explicit delimiters or 'ignore' instructions for the data ingested from the website.
Capability inventory: The skill utilizes the Read and Bash tools to execute the CLI binary.
Sanitization: The skill does not describe specific sanitization or filtering of the content crawled from the external service before it is processed by the agent.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands through the Bash tool to interact with the installed crestron-pp-cli and crestron-pp-mcp binaries.

pp-crestron