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 highly actionable body: concrete, copy-paste-ready bm_* calls, situation-based tool routing, and a clearly numbered workflow reinforced by scenario playbooks and a confirm-before-acting checkpoint. Its main weakness is conciseness — repeated project-scoping guidance and self-evident timeframe rows — plus a missing recovery loop for empty results.
Suggestions
State project-scoping rules once: merge 'Project Discovery' and the 'Project scope' section (both explain bm_project(command="list") and the qualified-URL-vs-project point) and drop the redundant restatements in Guidelines #1 and #4.
Trim the Timeframe Reference table to the non-obvious accepted syntax (compact forms like "3d"/"7d", and that natural-language strings are accepted) and remove self-evident rows like '"today" | Current day'.
Add a short error-recovery step for empty or no-match results (e.g., what to do when bm_search returns nothing or bm_context yields an empty graph), turning the 'recovering from a refusal' mention into concrete guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient tool-specific detail, but padded in places: the Timeframe table spells out self-evident rows ('"today" | Current day') and project-scoping rules are repeated across 'Project scope', 'Project Discovery', and Guidelines #1/#4. Not 4 because this duplication is more than minor trimming; not 2 because there is no explanation of concepts Claude already knows and the bm_* API detail genuinely earns its tokens. | 3 / 5 |
Actionability | Every step ships copy-paste-ready calls with real parameter values (bm_context(identifier=..., depth=2, since="7d"), bm_search(query=..., compact=True), bm_read(..., start_line=120, end_line=180, section="Status")), and the playbooks cover the common routing cases. Not 4 because nothing is pseudocode — placeholders like '<project>' are appropriate parameterization, not gaps. | 5 / 5 |
Workflow Clarity | 'Building Context' is a clearly numbered 1–4 sequence with situation-based tool routing, scenario playbooks restate it per case, and Guidelines #5 ('Verify the reconstructed context is what the user actually needs before acting on it') plus 'If it's unclear, ask' act as checkpoints. Not 5 because there is no explicit error-recovery loop — 'recovering from a refusal' is name-checked but never elaborated, and empty search/context results have no follow-up guidance. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ absent) and the body is well-sectioned — scope, When to Use, Building Context, Memory URL Reference, Timeframe Reference, Scenario Playbooks, Project Discovery, Guidelines — making it easy to navigate. Not 5 because at ~175 lines the single file repeats scoping content across sections ('Project scope' vs 'Project Discovery' vs Guidelines) where a leaner split would improve navigation. | 4 / 5 |
Total | 16 / 20 Passed |