The Agent Skills Directory

[DATA_EXFILTRATION]: The skill provides instructions to access sensitive local files containing credentials and private keys. Specifically, it reads from ~/.git-credentials to extract GitHub tokens and accesses ~/.ssh/id_rsa (a private SSH key) to upload it as a GitHub repository secret. While these actions are tied to the skill's purpose of repository management, accessing such files is a high-risk activity that exposes credentials to the agent context.
[PROMPT_INJECTION]: The skill processes untrusted data from external sources (the GitHub REST API), including repository names, descriptions, issue titles, and pull request bodies. These data points are interpolated into commands and printed to the user without sanitization or protective boundary markers, creating a surface for indirect prompt injection.
Ingestion points: Ingests repository metadata, PR content, issue details, and workflow logs via gh api and curl commands.
Boundary markers: None provided; instructions do not advise the agent to ignore potentially malicious content within the ingested data.
Capability inventory: The skill has significant capabilities, including file system access, repository creation/modification, and secret management.
Sanitization: No evidence of data sanitization or escaping before interpolation into shell commands or display.
[COMMAND_EXECUTION]: The skill frequently uses python3 -c to execute small Python scripts for parsing JSON responses from the GitHub API. This involves dynamic code generation and execution within the local shell environment to process network data.

github-repo-management