Content
63%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.
The body is exceptionally actionable, with executable commands and examples for every feature, and destructive operations are properly guarded with dry-run previews. Its main weaknesses are token efficiency — a large amount of per-agent and embedded-runtime implementation detail is inlined — and the absence of any progressive-disclosure bundle structure to move that detail into reference files.
Suggestions
Move the per-agent registry details (~90 lines of agent-specific quirks and external guide links) and the embedded-hosts API semantics into reference files, keeping only the default-agent rule and a few common agents in SKILL.md.
Trim or remove the 30+-bullet "Core capabilities" list at the top, since nearly every bullet duplicates a dedicated section later in the file.
Convert release-note-style behavioral prose (e.g., the -32603 session-recovery rules and adapter-range migration notes) into a short "Behavior notes" or "Migration/deprecated" reference file, keeping only operator-relevant rules inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~720-line body inlines deep implementation internals (e.g., "may be replaced after an ACP internal load/resume error (`-32603`)"), release-note-style migration rules, and a 30+-bullet capabilities list that duplicates later sections, making it noticeably verbose for a SKILL.md even though the content is accurate and free of known-concept padding. | 2 / 5 |
Actionability | Commands are fully executable and copy-paste ready throughout: install ("npm i -g acpx"), per-subcommand examples, a jq NDJSON automation pipeline, prune recipes, and a complete runnable defineFlow TypeScript example covering the common cases. | 5 / 5 |
Workflow Clarity | Key multi-step workflows are clearly sequenced with validation checkpoints where it matters — destructive prune is demonstrated with "--dry-run" preview first, and the system-prompt override flow gives an explicit create-then-select sequence — though most other sections describe behavior rather than prescriptive validated sequences. | 4 / 5 |
Progressive Disclosure | Section headers are well-organized and external references are clearly signaled (e.g., "see the [OpenClaw guide](...)", "See `docs/flows.md`"), but no bundle files exist and content that clearly belongs in separate files — per-agent registry quirks for ~25 agents and embedded-host API semantics — is inlined in a single monolithic file. | 3 / 5 |
Total | 14 / 20 Passed |