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.
Dense, highly actionable code templates give this skill real implementation value with almost no concept over-explanation, but everything is inlined in one large file and the implementer's validation workflow (especially compensation testing) stays implicit. The generic Instructions/Do-not-use boilerplate and a few undefined symbols in the templates are the main trim-and-fix items. Anti-drift check against neighboring anchors confirmed all four placements.
Suggestions
Move the four code templates into separate bundle files (e.g. templates/orchestrator.py, templates/order-fulfillment.py, templates/choreography.py, templates/timeouts.py) and keep SKILL.md as a concise overview that signals when to use each, reducing the main body to selection guidance.
Add an explicit implementation workflow with validation checkpoints, e.g. 1) define steps and their compensations, 2) implement service handlers, 3) verify each compensation reverses its action, 4) inject failure at every step and confirm rollback — turning the "Don't skip compensation testing" advice into a concrete sequence with a feedback loop.
Fix the minor executability gaps: import timedelta and add a timeout_at field to SagaStep in Template 4, and define or footnote InsufficientInventoryError in Template 2; also delete the generic "Instructions" and "Do not use this skill when" boilerplate sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by dense, non-redundant code templates with no tutorial-style explanation of concepts Claude already knows, and Best Practices are terse ("Make steps idempotent - Safe to retry"). Not 5 because boilerplate sections — "Clarify goals, constraints, and required inputs" / "The task is unrelated to saga orchestration" — are generic filler that earns no tokens; not 3 because the unnecessary prose is a small fraction of an otherwise efficient ~490 lines. | 4 / 5 |
Actionability | Four largely complete Python templates (orchestrator base, order fulfillment, choreography, timeouts) provide near copy-paste-ready guidance covering the common cases. Not 5 because of minor executability gaps: Template 4 uses timedelta and step.timeout_at without importing timedelta or adding the field to SagaStep, InsufficientInventoryError is undefined, and saga_store/event_publisher interfaces are unspecified; not 3 because the code is real executable Python, not pseudocode. | 4 / 5 |
Workflow Clarity | The saga lifecycle sequence is modeled in code (start → execute → handle completion/failure → compensate) and the state table, but implementer-facing checkpoints are absent — "Don't skip compensation testing" is asserted with no verify/test workflow, and sagas involve rollback of distributed transactions, so missing validation steps cap this at 3. Not 4 because validation is implicit only; not 2 because a coherent sequence is clearly present. | 3 / 5 |
Progressive Disclosure | Sections are clearly headed (Core Concepts, Templates, Best Practices, Resources) and there are no nested references, but ~400 lines of code templates are inlined in SKILL.md with no references/ or scripts/ bundle files — content that clearly belongs in separate template files lives inline. Not 4 because for a skill this size the all-inline structure is a real organization gap; not 2 because the section structure is good, not minimal. | 3 / 5 |
Total | 14 / 20 Passed |