The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/fetch_benchmark_repos.py utilizes subprocess.run to execute git clone commands for retrieving benchmark repositories. The command uses an argument list, which is a secure practice that avoids shell injection.
[EXTERNAL_DOWNLOADS]: The skill provides functionality to clone external repositories via Git for analysis. These operations are part of the intended benchmarking workflow and rely on URLs provided by the user via a manifest file.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external repositories. It includes defensive instructions that advise the agent to treat repository claims as leads and to prioritize the user's narrative, reducing the risk that the agent will be influenced by malicious content in analyzed repositories.

academic-repo-analyzer