Content
57%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.
This is a highly actionable, bash-first skill with concrete executable commands for every major workflow, undermined by padding, repetition, dated notes, and a monolithic single-file structure. The biggest structural gaps are the absence of validation checkpoints in batch/destructive workflows and the lack of any reference-file split.
Suggestions
Add validation steps before posting/creating PRs in the batch review and parallel issue-fixing workflows (e.g. verify tests pass or diff reviews cleanly before `gh pr create`/`gh pr comment`).
Split per-agent sections (Codex flags, Pi provider options) and the advanced batch/worktree playbooks into reference files (e.g. references/codex.md, references/parallel-workflows.md), keeping SKILL.md a concise overview.
Trim repetition and fluff: state the PTY requirement once, move dated items ("Learnings (Jan 2026)", "PR #584") into a notes section, and cut casual asides; fix the missing `exec` in the `codex --yolo` background example.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — parameter/action tables and command examples carry real weight — but includes unnecessary padding: repeated PTY warnings ("Always use pty:true", the ⚠️ section, Rules #1, and a Learnings bullet all restate it), casual asides ("like your soul.md 😅", "Parallel army!", "Sass works"), and time-sensitive details ("Learnings (Jan 2026)", "PR #584, merged Jan 2026", "gpt-5.2-codex default") that are not isolated in a notes/deprecated section. This matches anchor 3 (mostly efficient, some unnecessary explanation, could be tightened) rather than 4, which requires only minor trimmable instances. | 3 / 5 |
Actionability | Nearly all guidance is concrete, executable bash with the exact harness syntax ("bash pty:true workdir:~/project background:true command:...", "process action:log sessionId:XXX") covering one-shots, background monitoring, PR review, batch reviews, and worktree workflows. Minor gaps keep it from 5: placeholder templates ("gh pr comment <PR#> --body \"<review content>\"", "Fix issue #78: <description>") and an inconsistent invocation ("codex --yolo 'Refactor the auth module'" at line 119 lacks the "exec" subcommand used everywhere else). | 4 / 5 |
Workflow Clarity | Multi-step sequences are clearly laid out (start → monitor → poll → submit → kill; the 5-step numbered worktree workflow with cleanup), but batch operations lack validation checkpoints: the batch PR review flow goes straight from "Monitor all" to "Post results to GitHub" without checking output quality, and the parallel issue-fixing flow creates PRs after fixes without verifying tests/builds pass. Per the rubric, missing validation in batch workflows caps workflow_clarity at 3 — it is above 2 because sequences are well-defined with monitoring steps, but cannot reach 4 without outcome verification. | 3 / 5 |
Progressive Disclosure | The single SKILL.md (~285 lines) has clear section headers but inlines content that belongs in separate reference files — per-tool guides (Codex flags, Pi providers), batch PR review patterns, and the parallel worktree playbook could each be one-level-deep references, keeping SKILL.md as an overview. No bundle files exist (no references/, scripts/, or assets/), so the structure is a well-headed but monolithic inline document, matching anchor 3 (some structure, content that should be separate is inline) rather than 4 (most content appropriately placed with references). | 3 / 5 |
Total | 13 / 20 Passed |