Content
81%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, executable skill body with a clear approval-gated workflow, built-in validation, and strong error-recovery guidance. Its weaknesses are moderate padding (a marketing/roadmap section and guardrails duplicated across two sections) and missed opportunity to split the full API surface and troubleshooting detail into reference files.
Suggestions
Remove or trim the 'Why this lives in Bankr Skills' paragraph — the roadmap information ('Bankr-settled transactions and agent-economy hooks are on the roadmap') is not actionable for any current task and adds token cost to every invocation.
Merge 'Notes & guardrails' into 'Safety & approval' (or delete it) — the host-level-execution, preview-then-install, and no-credentials points repeat guidance already stated in the earlier section, which could be recapped in one line.
Move the full export list ('fetchRelease, parseChecksums, selectAsset, getTarget, isNewer' plus the type names) and the troubleshooting table into a references/ file (e.g. references/api.md, references/troubleshooting.md), keeping only the two or three most common failure modes in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient tables and commands, but there is unnecessary content: the "Why this lives in Bankr Skills" paragraph about the agent-economy roadmap ("Bankr-settled transactions and agent-economy hooks are **on the roadmap**") tells Claude nothing actionable, and the guardrails are stated twice — "Never trigger one as a silent side effect" / "never automatically from an agent workflow" appear in both "Safety & approval" and "Notes & guardrails". Not 4: the padding is more than minor instances (a full marketing/roadmap section plus duplicated sections); not 2: the bulk is lean, actionable tables and commands. | 3 / 5 |
Actionability | Copy-paste ready throughout: "npm install -g hermesone@0.2.0", "hermesone plan", "hermesone install", a complete command table, a flag/env table, an executable TypeScript example with the exact shape of plan() output ("{ tag, asset, url, size, sha512 }") and progress phases, and a troubleshooting table keyed to exact error strings. Not below 5: common cases (install, pin, update, programmatic use, failure modes) all have specific runnable examples. | 5 / 5 |
Workflow Clarity | The recommended flow is a clearly sequenced install → plan → install-on-confirmation pattern with the plan step serving as an explicit validation checkpoint for agent contexts; the sha512 verification (fail-closed, "the file was deleted and nothing is installed") is a built-in validation step, and the troubleshooting table provides feedback loops for error recovery (checksum mismatch, 403 rate limit, missing asset). Not below 5: for this host-modifying operation the sequence, checkpoints, and recovery paths are all explicit. | 5 / 5 |
Progressive Disclosure | Well-organized single-file body with clear section headers (Recommended flow, CLI reference, Pin, Checksum verification, Programmatic API, Troubleshooting) and internal anchor links, with no nested references — but there are no bundle files at all, and at ~140 lines the troubleshooting table and the full export list ("fetchRelease, parseChecksums, selectAsset...") could live in a references/ file. Not below 4: what is inline is navigable and appropriately signaled, not a monolithic wall; not 5: the doc is over 50 lines and some reference-grade material (full API surface, troubleshooting) is inlined rather than split. | 4 / 5 |
Total | 17 / 20 Passed |