Content
65%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 dense, example-rich reference that teaches Inngest's step model effectively through executable code and gotchas, and it is notably lean for its coverage. Its weaknesses are the total absence of validation/verification guidance for its batch and parallel patterns, and a monolithic structure whose only external reference is broken. Fixing the expressions.md reference and adding error/timeout checks to the pattern examples would lift the two lowest dimensions.
Suggestions
Add validation checkpoints to the batch/parallel patterns: check for null waitForEvent returns is covered, but extend it to verifying step outputs before combining chunk results and handling partial failures in Promise.all.
Fix the dangling reference: either create references/expressions.md at the linked path or remove/repair the link, and point the "inngest-flow-control" mention at its actual file path.
Consider moving the per-method API detail into a reference file so SKILL.md stays an overview, which would also let you demonstrate step.ai properly.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is code-first and assumes competence — "Each step is a separate HTTP request that can be independently retried and monitored" plus tight examples, with no explanation of concepts Claude already knows. Minor trimming opportunities exist: DO/DON'T lists restate nearby prose, emoji labels add tokens, and the "Remember: Steps make your functions durable... Embrace them!" closer earns nothing. Not 3, since padding is minor rather than a pattern. | 4 / 5 |
Actionability | Nearly every section ships concrete, executable TypeScript (step.run, waitForEvent with match/if, Promise.all parallelism, pagination loop), matching "concrete code with minor gaps". It falls short of 5 because `chunkArray` is an undefined helper, sending signals is only sketched as a comment ("// POST /v1/events with signal matching taskId"), and step.ai is claimed in the description and use cases but never demonstrated. | 4 / 5 |
Workflow Clarity | Sections are clearly organized with WRONG/CORRECT contrasts and a gotchas recap, but no validation or verification checkpoints appear anywhere. The Patterns section involves batch operations (pagination loops, chunking, parallel Promise.all), and the rubric states missing validation in batch-operation workflows caps workflow clarity at 3 — it cannot reach 4 despite good sequencing. | 3 / 5 |
Progressive Disclosure | Everything lives in one ~340-line SKILL.md that reads partly as an inline API reference, and the single reference link "See [Expression Syntax Reference](../references/expressions.md)" points to a nonexistent path (no references/ directory exists in the bundle). "See inngest-flow-control" is likewise invoked without a path. This is "some structure, references present but not clearly signaled / content that should be separate is inline"; it does not reach 4 because a referenced file is dangling and the flow-control pointer is unresolvable. | 3 / 5 |
Total | 14 / 20 Passed |