Content
75%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 well-structured, high-signal instructional skill: every step is concrete, non-obvious project rules (config precedence, session-identity, typed PR fields) are captured compactly, and validation guards are explicit. The main residual gaps are the absence of a worked note example and a slightly dense single-file layout.
Suggestions
Add one short worked example of a finished checkpoint note body (sections plus `[next_step]`/`[decision]` observation lines) so the observation format is copy-paste unambiguous.
Trim the opening paragraph's restatement of the description's trigger conditions to save tokens without losing guidance.
Consider moving the coding-profile field schema and config-precedence rules into a single one-level-deep reference file, keeping SKILL.md as a leaner overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense, imperative, and assumes competence — e.g. "Apply the `bm-writing` skill before drafting the note" and "Begin the body with `# <exact note title>`" — with no explanation of concepts Claude already knows. Minor trim opportunities remain: the opening paragraph re-states the description's trigger conditions, and the config-precedence paragraph is wordier than strictly needed. | 4 / 5 |
Actionability | Guidance is highly concrete for an instruction-only skill: exact commands (`git status --short`, `gh pr view --json ...`), a complete field schema (`note_type: session`, `tags: ["claude", "checkpoint"]`, `pull_request_number: "123"`), a section template, and one relation example (`- relates_to [[Exact existing note title]]`). Minor gaps: no worked example of a full note body and no example line showing the observation format (e.g. `[next_step] ...`), so it is mostly rather than fully executable. | 4 / 5 |
Workflow Clarity | A clear Gather -> Write -> Confirm sequence with explicit validation guards ("Do not claim a test passed unless you ran it", "Stop if the required coding fields cannot be proven", "never infer one" for session IDs) and a checklist of evidence to gather. It stops short of anchor 5 because there is no post-write verification or error-recovery loop (e.g. handling a failed `write_note` call), though the operation is not destructive or batch. | 4 / 5 |
Progressive Disclosure | No bundle files exist, and the single file is well organized under `## Gather`, `## Write`, `## Confirm` with clearly signaled field lists per profile. It is not anchor 5 because at ~127 lines the coding-profile field schema and config-precedence rules are borderline content that could live in a one-level-deep reference file, but nothing is buried or nested. | 4 / 5 |
Total | 16 / 20 Passed |