The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates external data from GitHub into the agent's context without sanitization.
Ingestion points: The skill reads pull request descriptions, labels, and issue details from GitHub using the gh command-line tool.
Boundary markers: No delimiters or safety instructions are provided to distinguish between system instructions and data retrieved from the repository.
Capability inventory: The agent has access to gh cli (read) and curl (write to webhooks), which could be leveraged by injected instructions to exfiltrate data or perform unauthorized actions.
Sanitization: The skill lacks logic to validate or escape external content before processing.
[DATA_EXFILTRATION]: The skill provides instructions to configure a Discord webhook for automatic updates. While user-controlled, this allows internal code changes and deployment details to be transmitted to an external service.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands using gh cli for repository interaction and curl for network requests.

changelog