Content
82%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.
A lean, highly actionable rebase runbook with real validation checkpoints around its destructive operations (force-push, history rewrite). The main defects are a workflow numbering error (step 4 is skipped) and the cherry-pick path being described rather than scripted.
Suggestions
Fix the workflow numbering: the steps jump from 3 to 5 with no step 4, which makes the sequence appear broken even though the content flows.
Provide executable commands for the cherry-pick strategy (e.g. `git checkout -b "${subject}__cherrypick__" "${target_ref}"` followed by the explicit cherry-pick invocation) so all three strategy branches are copy-paste ready.
Guard the post-push tracking step: `git branch --set-upstream-to="origin/${subject}"` fails if the remote branch does not yet exist, so cover the first-push case (e.g. rely on `git push -u` alone when `origin/${subject}` is absent).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is almost entirely executable commands plus genuinely non-obvious domain guidance (rerere recovery, filtered machine-state files, the worktree fetch caveat), with no padding and no explanation of concepts Claude already knows. It matches anchor 5 (lean, every token earns its place); anchor 4 would require some trimmable over-explanation, which is absent. | 5 / 5 |
Actionability | Context gathering, divergence analysis, simple/onto rebase, and push are given as exact runnable commands. However the cherry-pick strategy is only described ("create `<subject>__cherrypick__` from `target_ref` and cherry-pick subject commits oldest-first") without executable commands, and `git branch --set-upstream-to="origin/${subject}"` silently assumes the remote branch already exists. This is anchor 4 (mostly executable, minor gaps) rather than anchor 5 (copy-paste ready across all paths). | 4 / 5 |
Workflow Clarity | The sequence has strong checkpoints for a destructive workflow: dirty-tree stop condition, analysis shown before proceeding, conflict-marker verification before `--continue`, rerere-based recovery, and `--force-with-lease` for the push. But the numbered workflow skips step 4 entirely (it jumps from 3 to 5), a real sequencing defect that keeps it below anchor 5; it stays above anchor 3 because validation for the destructive and batch operations is explicitly present. | 4 / 5 |
Progressive Disclosure | The skill is a single well-organized file with no bundle files, and the Workflow sectioning is clear. At roughly 75 lines it exceeds the simple-skill threshold, and the strategy-selection table and conflict-resolution heuristics are cohesive but arguably separable into a reference file. This lands on anchor 4 (good structure, minor organization gaps) rather than anchor 5 (clean overview with well-signaled split references). | 4 / 5 |
Total | 17 / 20 Passed |