The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to simulate GitHub lifecycle events (such as pull requests, issues, and comments) by injecting data into a local development server. This data is then processed by the agent or the application under test, creating an attack surface where malicious instructions could be embedded in simulated content.
Ingestion points: Data enters the agent's context via JSON payloads and proxy overlays fetched from http://127.0.0.1:7777 using curl.
Boundary markers: No specific delimiters or safety warnings are implemented to distinguish between simulated data and system instructions.
Capability inventory: The skill utilizes shell execution for curl, jq, and mise commands to interact with the local environment.
Sanitization: There is no evidence of sanitization or escaping of the simulated data before it is processed or displayed.
[COMMAND_EXECUTION]: The skill uses several shell commands (curl, jq, mise, and git) to manage a local development server and verify the state of the workspace. These operations are limited to the loopback interface (127.0.0.1) and local repository metadata, which is consistent with its stated purpose as a development utility.

desktop-devserver