The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill automatically manages its own dependencies by running bun install when the tools are invoked. This fetches packages from the NPM registry using the Bun package manager.
[COMMAND_EXECUTION]: The skill utilizes several internal scripts and system utilities (git, gh, xcrun) to perform its stated functions, such as auditing worktrees, managing pull requests, and orchestrating subagents.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including GitHub pull request comments and session transcripts. It implements proactive security measures, such as explicit instructions to the agent to treat comments as untrusted and data sanitization (e.g., escaping TSV/spreadsheet injection characters) in its storage layer.
[DATA_EXPOSURE]: The skill is designed to read session transcripts to facilitate task resumption and performance evaluation. It includes explicit instructions to the agent to respect workspace boundaries and avoid reading transcripts from unrelated projects.

poteto-mode