Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-crafted orchestration skill with excellent workflow clarity and actionability. The multi-step process is clearly sequenced with explicit validation, error handling, and user interaction points at every stage. The main weakness is that the skill is somewhat long for a sequencer that claims to delegate logic to sub-skills — some sections (worktree cleanup, feature tracker updates) are detailed enough to warrant extraction to reference files.
Suggestions
Extract the worktree cleanup bash script and logic to a reference file (e.g., references/worktree-cleanup.md) to reduce the main skill's length and reinforce the sequencer pattern.
Consider extracting the Feature Tracker Update section (Step 4) to a reference file, as it contains detailed logic that could be reused or maintained independently.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is fairly long but most content is necessary for the complex multi-step orchestration workflow. Some sections could be tightened (e.g., the worktree cleanup explanation, the unmerged worktree note), and there's mild over-explanation in places like the CHANGE_RECORD scanning criteria, but overall it respects Claude's intelligence and doesn't explain basic concepts. | 2 / 3 |
Actionability | The skill provides concrete, executable commands throughout (gh pr merge, git checkout, worktree cleanup bash scripts), specific CLI invocations for both GitHub and Bitbucket platforms, exact file paths to read, and precise data structures to extract. The AskUserQuestion prompts include specific option text. Code blocks are copy-paste ready. | 3 / 3 |
Workflow Clarity | The workflow is exceptionally well-sequenced with clear numbered steps, explicit validation checkpoints (platform gate, PR state verification, merge failure handling), feedback loops (re-run analyzer after each merge, retry on failure), and clear branching logic for each user choice. Error recovery is explicit at every stage — failed merges skip and continue, failed analysis offers retry, worktree cleanup is conditional on success. | 3 / 3 |
Progressive Disclosure | The skill references external files appropriately (conflict-classification.md, step-0-fast-path.md, ensure-config.md, sub-skills like review-pr and batch-pr-analyzer), but the SKILL.md itself is quite long and monolithic. The worktree cleanup bash block and feature tracker update logic could potentially be extracted to reference files. No bundle files were provided to verify referenced paths exist, but the references are one-level deep and clearly signaled. | 2 / 3 |
Total | 10 / 12 Passed |