Content
77%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 highly actionable, well-sequenced operational skill with excellent command-level specificity and validation checkpoints. Its weaknesses are structural: the two bundled reference files are orphaned (never referenced from the body, which duplicates their content inline), the ledger script is inlined rather than placed in scripts/, and the core safety constraint is repeated several times.
Suggestions
Link references/bugbot-triage.md and references/merge-safety.md from the body (e.g., in step 3's review lane and in Rules) and trim the inline duplicates, so the bundle's one-level-deep references are actually discoverable.
Move the step-4 ledger-writing node script into scripts/ (e.g., scripts/append-ledger.sh) and invoke it with the PR number, head SHA, and evidence as arguments, shrinking SKILL.md substantially.
State the never-merge/rebase/force-push constraint once in Rules and refer back to it from step 3 and the verdict policy instead of repeating it nearly verbatim four times.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with genuinely non-obvious operational knowledge (gh exit-code quirks, `reviewDecision` empty-string semantics, `HAS_HOOKS` classification), but the never-merge/rebase/force-push constraint is repeated four times (intro, step 3, the verdict policy note, and Rules), and step 4 embeds a long inline node script that belongs in scripts/. It is mostly efficient but could be tightened, matching anchor 3 rather than 4. | 3 / 5 |
Actionability | Every step carries exact executable commands (`gh pr view <N> --json state,mergeable,...`, `gh pr checks <N> --json name,bucket`, the branch-protection `gh api` call, the copy-paste-ready ledger node script with explicit placeholders) plus fully specified classification rules with field values and edge cases (bucket `cancel`, 60-second no-checks window, `HAS_HOOKS`). This matches anchor 5 — copy-paste ready commands covering the common cases — and not anchor 4, which expects minor gaps. | 5 / 5 |
Workflow Clarity | The Loop is a clearly numbered sequence (read state → apply policy → act → ledger → report) with explicit validation checkpoints ("verify the checkout sits on that PR's head before mutating anything", "Confirm the first read reports the requested PR") and feedback loops (re-poll on UNKNOWN, "wait for its push before re-classifying", the no-checks confirmation window), with explicit stop conditions. This matches anchor 5; anchor 4's 'minor validation gaps' does not apply. | 5 / 5 |
Progressive Disclosure | The bundle contains references/bugbot-triage.md and references/merge-safety.md, but the body never links to or signals either file — instead it inlines condensed duplicates of both (bot-triage rules in step 3 and Rules; merge-safety rules in the intro and Rules), and the ledger script that belongs in scripts/ is inlined. This matches anchor 3 ('references present but not clearly signaled; content that should be separate is inline') — not 4, since the references are not navigable from SKILL.md at all, and not 2, since the body itself has clear section structure. | 3 / 5 |
Total | 16 / 20 Passed |