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 highly actionable, with exact paths, required params, and runnable curl examples covering the common cases, and multi-step workflows are well sequenced. It is weakened by verbose dated pricing narrative, two internal contradictions (stale $0.02 cost math and the blockrun_surf_* tool-name instruction), and a monolithic 250-line catalog that belongs in a reference file.
Suggestions
Move the 83-endpoint catalog into a references/ file (e.g. references/endpoints.md) and keep SKILL.md as an overview with the 'when to use' examples and a table of domains, so the catalog loads only when needed.
Trim the pricing section to the current flat rates plus the 'trust the 402 quote over the table' rule, and relocate the superseded $0.001/$0.005/$0.020 tier history and the 2026-09-05 verification narrative into a clearly labeled 'deprecated pricing' note.
Fix the internal contradictions: update the example-flow cost arithmetic to the measured $0.0075–$0.0085 per-call rates, and remove or correct the 'Use the OpenClaw tool name blockrun_surf_*' instruction that contradicts the 'No typed blockrun_surf_* tools are registered' statement.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The endpoint catalog tables and curl examples are dense and efficient, but the pricing section carries dated narrative that could be tightened — 'Surf used to be tiered at $0.001 / $0.005 / $0.020. It is not any more', 'Verified 2026-09-05 by measurement, not from a price page', and the digression about when a Solana/Base price difference 'would be the bug worth reporting'. The rubric penalizes time-sensitive information (specific verification dates, superseded tier prices) that is not placed in a deprecated/old-patterns section. | 3 / 5 |
Actionability | Guidance is largely copy-paste executable — full curl commands with real hosts, paths, and JSON bodies, exact required params per endpoint, and documented 400 pre-check behavior. Minor gaps: the example-flow cost line ('1 × $0.02 ... = $0.04 total') contradicts the flat $0.0075/$0.0085 pricing table, and line 277 instructs using tool name 'blockrun_surf_*' while line 274 says no such tools are registered — a stale/contradictory pair that leaves the agent unsure which figure or invocation path to trust. | 4 / 5 |
Workflow Clarity | Multi-step flows are clearly sequenced: schema-then-SQL with 'cache it locally', search-then-mindshare with the correct param names called out, and the wallet-detail follow-up decision points. The pre-settlement 400 validation behavior is documented ('you get 400 { missing_params, all_required, docs } and you are NOT charged'), but there are no explicit error-recovery loops (what to do on timeout or unexpected 402 amounts), keeping it just below the top anchor. | 4 / 5 |
Progressive Disclosure | Sections and tables are well organized, but the skill is a ~250-line monolith with no bundle files at all: the entire 83-endpoint catalog, per-domain param tables, and the multi-paragraph pricing-rail discussion are all inlined in SKILL.md. The rubric's structure expects an overview pointing to one-level-deep reference files for bulk API material, so the catalog 'content that should be separate is inline' — it has structure but the split into a references/ file is missing. | 3 / 5 |
Total | 14 / 20 Passed |