Content
71%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, highly actionable internal skill: every section carries project-specific knowledge with executable code, exact paths, and a concrete debugging checklist. Its weaknesses are repetition of the core enqueue-then-dispatch rule across four sections and a monolithic single-file layout where the SQL schema, Netlify specifics, and channel-settings rules could be split into reference files.
Suggestions
Consolidate the recurring statement of the core rule ('enqueue to SQL, return 200, process in a fresh execution, sweep for recovery') so it is stated fully once in 'Rule' and only referenced from 'Why', 'Cross-Platform Considerations', and 'Why Fire-and-Forget on Serverless Is Unreliable'.
Move the `integration_pending_tasks` SQL DDL and the Netlify durable-dispatch rollout details (flags, scopes, lease cutoffs) into a references/ file (e.g. references/pending-tasks-schema.md and references/durable-dispatch.md), keeping only the claim semantics and default behavior in SKILL.md.
Add an explicit verify step to 'Adding a New Platform Adapter' (e.g. 'run the catalog and adapter tests, then send a signed test webhook and confirm the task row reaches completed') so the sequence has a validation checkpoint like the debugging checklist already does.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly project-specific, non-obvious knowledge (host freeze behavior, atomic claims, retry cutoffs) that earns its tokens, but the core 'persist, then self-fire, with a recovery sweep' rule is restated across 'Rule', 'Why', 'Cross-Platform Considerations', and 'Why Fire-and-Forget on Serverless Is Unreliable', and the 'Netlify durable recovery' subsection repeats lease/cutoff details twice. This is more than anchor 4's 'minor instances of over-explanation' — several sections could be consolidated — but it is not the padding of anchors 1-2. | 3 / 5 |
Actionability | Fully executable guidance throughout: a complete `PlatformAdapter` TypeScript template with all required methods, the exact `integration_pending_tasks` SQL DDL, real file paths (`packages/core/src/integrations/webhook-handler.ts`), concrete env vars (`AGENT_INTEGRATION_DURABLE_DISPATCH`, `WEBHOOK_BASE_URL`), a copy-paste debug query (`SELECT * FROM integration_pending_tasks WHERE external_thread_id = '...'`), and named functions (`claimPendingTask`, `setMountedChannels`). Matches anchor 5's copy-paste-ready coverage of common cases. | 5 / 5 |
Workflow Clarity | The main flow is a numbered 9-step diagram with a recovery sweep, and 'Adding a New Platform Adapter' is a clear 4-step sequence with a built-in validation gate ('catalog.spec.ts fails until the catalog and BUILT_IN_INTEGRATION_ADAPTER_IDS match') plus a 6-item debugging checklist with error-recovery guidance. It falls short of anchor 5 because the adapter-creation steps themselves lack an explicit verify/test checkpoint after registration, and the checklist is diagnostic rather than a validate→fix→retry loop inside the workflow. | 4 / 5 |
Progressive Disclosure | Section headers, tables, and a 'Related Skills' list give the 390-line body real structure, but everything lives in one file: the SQL schema, the Netlify durable-dispatch rollout details, and the Settings › Channels rules are each substantial enough to be separate reference files, and no bundle files (references/, scripts/, assets/) exist to offload them. This matches anchor 3's 'content that should be separate is inline' with decent organization, not anchor 2's structureless inline dump. | 3 / 5 |
Total | 15 / 20 Passed |