Content
86%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 lean, well-structured orchestrator skill: a clear four-phase arc with per-phase delegation to named sibling skills, real CLI commands, sensible principles, and an explicit stop condition. The main gaps are high-level directives in the gather/plan phases and no built-in verification loop, since correctness checking is pushed entirely to the separate eval flow.
Suggestions
Add one concrete example of 'enough material' at the end of phase 1 (e.g., what a gathered inventory looks like: skill files, code pointers, transcripts) so 'Make sure you have enough before moving on' is checkable rather than judgment-only.
Show the actual scaffolding invocation in phase 3 (e.g., `tessl plugin new <name>` with its expected output layout) so the build step is copy-paste ready instead of delegating the syntax entirely to build-composition and `tessl --help`.
Add a short error-recovery note for the arc itself — what to do when gathering still leaves gaps after hunting, or when the user rejects the plan — so the workflow has an explicit feedback loop rather than only forward gates.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~30-line body is lean: every section carries task-specific instruction (phase delegation, CLI commands, principles, a stop condition) with no padding and no explanation of concepts Claude already knows. The eval point appears in both the intro and its own section, but the intro explicitly cross-references it ("see the end"), so it reads as deliberate signposting rather than waste. | 5 / 5 |
Actionability | Concrete guidance is present — named sub-skills per phase, real commands ("tessl skill new", "tessl plugin new", "Inspect tessl --help"), and the path convention "skills/<name>/SKILL.md" — but several directives stay high-level ("Make sure you have enough before moving on", "Produce a short plan and confirm it", "add the planned hook or MCP fields and companion files") without the specific steps or examples needed to execute them directly. Mostly executable with minor gaps, not the copy-paste-ready 5. | 4 / 5 |
Workflow Clarity | The four phases are numbered and sequenced, each with a named skill and purpose, gates exist ("Make sure you have enough before moving on", "confirm it", user agreement before changing content), and a "When to stop" section defines termination. It falls short of 5 because there is no explicit error-recovery or output-verification loop — verifying the built composition is deferred to the downstream eval flow rather than checked here. | 4 / 5 |
Progressive Disclosure | Under 50 lines with well-organized sections and no inlined content that belongs in separate files; the only references are one-level-deep, clearly signaled pointers to sibling skills (gather-context, plan-composition, build-composition, publish-plugin, tessl/skill-optimizer) and no bundle files exist to restructure. Per the simple-skill guideline, this scores 5. | 5 / 5 |
Total | 18 / 20 Passed |