Content
78%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 tight, well-structured instruction skill: concrete commands, specific data sources, a complete output template, explicit error-recovery loops, and clean section organization with one-level references. The remaining gaps are the Contract section duplicating the Phases, unspecified tool calls for attendee-context and task loading, and no inline validation checkpoint before the briefing is compiled.
Suggestions
Merge or trim the Contract section — it restates the Phases almost verbatim; fold the guarantees into the phase list or output format.
Name the exact tool calls for phases 1 and 3 (e.g., which of search/query/get_page/get_timeline to use for attendee brain pages and loading `ops/tasks`).
Add an inline validation checkpoint before Phase 4, e.g. confirm every meeting has attendee context and `gbrain waiting` succeeded before compiling the briefing.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~49-line body is lean, assumes competence, and explains nothing Claude already knows, but the "Contract" section restates the four phases nearly verbatim ("Calendar/meetings for today are loaded with brain context per attendee" vs. Phase 1, "Open threads from yesterday are surfaced" vs. Phase 2) — minor duplication that could be trimmed. It is not 5 because of that redundancy and not 3 because every other section earns its tokens. | 4 / 5 |
Actionability | Concrete guidance is present: a specific command ("`gbrain waiting --json` is the real data source"), a named data source ("Load `ops/tasks` from brain"), and a full copy-paste output template. It is not 5 because phases 1 and 3 do not specify the actual tool invocations (which of search/query/get_page/get_timeline to use for attendee context or task loading), and not 3 because the guidance is mostly executable with specific commands and a complete output format. | 4 / 5 |
Workflow Clarity | The four numbered phases are clearly sequenced, and "When it fails" supplies genuine feedback loops ("run (or ask the user to run) the sync it names, then retry; do not prep from stale mail"). It is not 5 because the happy path lacks an inline validation checkpoint (e.g., confirming every meeting has attendee context before compiling the briefing), and not 3 because error-recovery checkpoints are explicit rather than missing. | 4 / 5 |
Progressive Disclosure | The body is under 50 lines with well-organized sections (Contract, Phases, Output Format, When it fails, Anti-Patterns), and its external pointers are one level deep and clearly signaled ("see `skills/google-loops/SKILL.md`" and the linked agent-operator-protocol). Per the simple-skill guideline this earns a 5 with no separate bundle files needed. | 5 / 5 |
Total | 17 / 20 Passed |