Content
46%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill body has a solid, check-pointed pre-flight section but collapses after it: Target Selection is empty, the actual revert logic is delegated to eleven sub-skill files that do not exist, and destructive-operation verification is missing. Boilerplate use/do-not-use sections pad the body without adding information.
Suggestions
Create the referenced `sub-skills/*.md` and `resources/implementation-playbook.md` files (or inline their content) — the body currently links to eleven files that are missing from the bundle, making the skill non-functional.
Fill in the empty "Target Selection" section with the actual revert steps (e.g. `git log` to locate the unit's commits, `git revert <range>` or `git reset`, verification with `git status`/`git log`) so the destructive core has executable instructions and a post-revert validation loop.
Delete the circular "Use this skill when" / "Do not use this skill when" sections and the generic Instructions bullets, replacing them with the real decision tree the sub-skill headings imply.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The pre-flight section is lean and useful, but the "Use this skill when" / "Do not use this skill when" sections are circular boilerplate ("Working on revert track tasks or workflows", "The task is unrelated to revert track") and the Instructions bullets ("Apply relevant best practices and validate outcomes") are generic filler — matching the anchor for mostly efficient with some unnecessary padding. Not 2 because there is no over-explanation of known concepts and most tokens do work. | 3 / 5 |
Actionability | Pre-flight checks give concrete, executable guidance ("Run `git status`", "Check `conductor/tracks.md` exists", a specific warning dialog with three options), but the "Target Selection" section is an empty heading and no actual revert commands or steps appear anywhere — the anchor for some concrete guidance that is incomplete. Not 4 because the core operation of the skill has zero executable instructions in the body; not 2 because the pre-flight portion is genuinely executable. | 3 / 5 |
Workflow Clarity | The pre-flight sequence is well ordered with explicit checkpoints (uncommitted-change halt with options, merge/rebase detection with "Halt and explain resolution steps"), but this is a destructive git operation with no post-revert verification or feedback loop, and the main workflow is a heading with no steps. Per the guideline capping destructive/batch skills without validation at 3, this cannot exceed 3; it is above 2 because a real, check-pointed pre-flight sequence is present. | 3 / 5 |
Progressive Disclosure | The body points to `resources/implementation-playbook.md` and eleven `./sub-skills/*.md` files, and none of these exist in the bundle — the navigation is a shell of headings linking to nothing, which is broken rather than merely under-organized. Scored against the actual bundle structure per the guidelines, this fits the anchor for minimal structure with unusable references (2); it is not 3 because references that resolve to nothing provide no navigable organization at all. | 2 / 5 |
Total | 11 / 20 Passed |