Review notes:
- This change correctly addresses the issue of `fzf --tmux` hanging in VSCode and Zed terminals by unsetting the `TMUX` environment variable.
- The conditional check for `$TERM_PROGRAM` is appropriate for identifying these specific environments.
- This fix improves the user experience when using `fzf` within these integrated terminals.

Flake lock file updates:

• Updated input 'jj-repo':
    'git+ssh://[email protected]/pze/jj.git?ref=refs/heads/main&rev=0a9ab49dc5e59d9d6032477d0f4e916a24c1ca21&shallow=1' (2025-06-13)
  → 'git+ssh://[email protected]/pze/jj.git?ref=refs/heads/main&rev=bc83eb0a4837bdec89792967dd28674141f4a364&shallow=1' (2025-06-23)

Review notes:
- The diff indicates a near-complete rewrite of the `gen-prompt.md` file.
- Ensure the new content accurately reflects the current prompt generation logic and requirements.
- Verify that all necessary information from the previous version has been either incorporated or intentionally removed.
- Consider if this extensive change warrants a more detailed commit message body explaining the rationale for the rewrite.

Review notes:
- The model name was updated from `claude-sonnet-4` to `claude-3.7-sonnet`. This is a minor version bump and should be compatible.
- Ensure that the new model version `claude-3.7-sonnet` is indeed the correct and intended model to use, as `claude-sonnet-4` might refer to a different, potentially newer, internal alias or a future model.
- Verify that this change does not negatively impact the prompt's performance or output quality.

Review notes:
- The previous logic could leave temporary files in /tmp if the user aborted the editor without saving content.
- The new approach ensures the temporary file is removed in both success and failure cases.
- The temporary file is now stored in a dedicated directory with a timestamp, which is good for debugging if issues arise before deletion.
- Consider using `mktemp -d` to create a temporary directory and then a file inside it, which might be cleaner for managing temporary resources.

Add a detailed prompt for LLM-based task plan reviews, including review process, report format, and guidelines.
Introduce a new `goose-review-plan` script to facilitate the review process.

Review notes:
- The prompt provides a comprehensive structure for task plan reviews, which should improve consistency and quality of LLM outputs.
- The `goose-review-plan` script integrates well with the existing `goose` CLI, making the new functionality easily accessible.
- Ensure the LLM is fine-tuned or capable enough to follow the detailed instructions and generate reports in the specified format.
- Consider adding examples of good and bad task plans to the prompt for better context.
- The script currently checks for argument count; ensure robust error handling for file existence and readability.

Review notes:
- The change correctly addresses the issue of `interactive_mode` not being passed to `goose` when it's an empty string.
- Using `set -l` and `eval` for command construction is a common pattern in fish shell for dynamic commands.
- The previous single-line command was brittle and would fail if `$interactive_mode` was empty, as it would pass an empty string as an argument to `goose`. The new approach correctly omits the `--interactive` flag when `$interactive_mode` is empty.