The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to obtain templates and installation scripts by cloning a community-maintained GitHub repository at https://github.com/JDDavenport/context-kit.
[COMMAND_EXECUTION]: The skill provides various shell commands for the agent to execute, including cloning repositories (git clone), inspecting script segments (sed), searching for files (find), and modifying project configuration (printf to .gitignore).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and adapt external Markdown artifacts that serve as personal context. These files represent a potential ingestion point for hidden instructions that could target the agent's behavior.
Ingestion points: Personal context artifacts like pca-wiki.md and pca-voice.md sourced from an external repository or provided by the user.
Boundary markers: The skill emphasizes manual oversight and user review but lacks technical delimiters or explicit system instructions to isolate and ignore embedded commands within the ingested files.
Capability inventory: The agent is authorized to use file system tools, network-based git commands, and local configuration management.
Sanitization: The skill recommends that the user manually trim templates and omit sensitive details, but it does not provide automated sanitization or validation of the content before it is processed.

context-kit