Content
92%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.
An exceptionally well-structured review skill: executable commands, a validated workflow with feedback loops, and model progressive disclosure (a 'when to read' table plus conditional-loading guidance, with all nine bundle files present, real, and consistent with the body's claims). The only weakness is mild redundancy around the set -e/-u stance, which keeps conciseness just below the top anchor.
Suggestions
State the set -e/set -u anti-pattern stance once in the intro (with the reference pointer) and have Steps 3b and --fix reference it by name instead of restating it.
Trim the opening framing paragraph ('grounded in named authorities rather than ad-hoc opinion') to one clause; the named-authority table and closing goal statement already carry this point.
Consider compressing the output skeleton's severity placeholder sections (### High/### Medium/### Low '...') into a single annotated placeholder to save ~10 lines without losing the template.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is procedural, assumes Claude's competence, and wastes nothing on shell tutorials. However, the set -e/set -u error-handling stance is restated in three places (intro, Step 3b, and the --fix ordering) and the opening framing paragraph could be trimmed — minor instances of over-explanation, matching the 4 anchor rather than the every-token-earns-its-place of 5. | 4 / 5 |
Actionability | Fully executable guidance throughout: concrete wrapper invocations (scripts/detect_shell.sh <file>, scripts/run_shellcheck.sh <file> [bash|sh]), exact fix commands (shfmt -w -i 2 -ci -bn -sr --, diff | patch -p1), a copy-paste output skeleton, and named ShellCheck codes (SC3045). All commands reference verified bundle scripts, matching the 5 anchor. | 5 / 5 |
Workflow Clarity | Steps 1–5 are clearly sequenced (dialect detection → static analysis → prioritized manual review → structured output → escalation), with explicit validation checkpoints: static analysis runs before and is re-run after --fix to confirm, --fix requires user confirmation before applying, and the escalation rule handles disagreement and runtime-dependent constructs — a genuine feedback loop matching the 5 anchor. | 5 / 5 |
Progressive Disclosure | A reference table with a 'When to read' column, explicit conditional-loading guidance ('Load only the file(s) relevant to the current script'), and one-level-deep references — all nine bundle files exist and their contents match the body's claims (e.g., the scripts' header comments confirm the zsh-via-bash and shfmt-flag behavior). Content is appropriately split with detail pushed to references, matching the 5 anchor. | 5 / 5 |
Total | 19 / 20 Passed |