The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several command-line tools including pip, python, git, and grep for environment management and code analysis. It also runs azpysdk, a utility provided by the vendor for Azure SDK development tasks. All commands are scoped to the package being processed and are run within a virtual environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core workflow of analyzing external, untrusted content.
Ingestion points: External data enters the agent's context through GitHub issue URLs (e.g., https://github.com/Azure/azure-sdk-for-python/issues/...) and package files (Python code, RST, and Markdown) analyzed during Steps 0, 3, and 5.
Boundary markers: Absent. There are no specific instructions or delimiters used to isolate or mark the ingested data as untrusted to prevent the agent from following embedded instructions.
Capability inventory: The skill possesses significant capabilities, including shell command execution (pip, python, git, grep, azpysdk), the ability to modify files in the repository (Step 7), and the ability to create GitHub pull requests (Step 10).
Sanitization: There is no evidence of sanitization or validation performed on the issue descriptions or file contents before they are processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill references and potentially downloads content from official domains such as azure.github.io, sphinx-doc.org, and the Azure/azure-sdk-for-python GitHub repository. These downloads are limited to documentation standards and developer tool usage guides consistent with the skill's purpose.

fix-sphinx