The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a local file named RPGSTUDIO.md in the current working directory to establish project context, including the BASE_URL. This creates a vulnerability surface where a malicious file placed in the user's workspace could override API targets or inject instructions.
Ingestion points: RPGSTUDIO.md file in the workspace.
Boundary markers: Absent; no specific delimiters or instructions are provided to the agent to distinguish between safe and untrusted content in the local file.
Capability inventory: The skill uses curl to perform network operations and has the ability to read and write to the local file system.
Sanitization: Absent; the skill does not perform validation or filtering on the values retrieved from the local context file.
[DATA_EXFILTRATION]: The skill uses the RPGSTUDIO_API_KEY environment variable in HTTP headers for requests made via curl. Because the destination BASE_URL can be overridden by a local file, there is a risk that the key could be sent to an attacker-controlled endpoint.
[COMMAND_EXECUTION]: The skill relies on executing curl commands to interact with the RPGJS Studio API. While necessary for its purpose, the reliance on shell commands with dynamically resolved URLs increases the risk of exfiltration if the environment is compromised.
[NO_CODE]: The skill package does not contain any standalone executable script files or binaries, relying solely on markdown instructions and system-level tools like curl.

rpgjs-studio