Content
78%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 strong, opinionated overview that establishes defaults, safety guardrails, and an operating procedure, with an exemplary progressive-disclosure structure verified against the real bundle. The main improvement levers are removing the description-duplicating trigger list, consolidating the repeated v1-default statements, and adding a couple of inline snippets for the routing-only task paths.
Suggestions
Cut or compress the 'What this Skill is for' section — it restates the frontmatter description almost verbatim; a one-line pointer to the trigger list would save ~15 lines.
State the transaction-v1-by-default rule once (section 1) and reference it from sections 2, 3, and the operating procedure instead of re-explaining it in each place.
Add a minimal inline code example for at least one routing-only path (e.g., a 5-line Anchor program skeleton or a token-creation call) so the most common program-layer tasks have executable guidance before deferring to references/programs/anchor.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational (tables, short bullets, executable snippets, no explaining of concepts Claude already knows), but has trimmable redundancy: the 15-line "What this Skill is for" list largely restates the frontmatter description, and the v1-by-default rule is repeated in sections 1, 2, 3, and the operating procedure. This fits 'efficient; minor instances of over-explanation that could be trimmed' rather than the level-5 'every token earns its place' anchor, and it is well above the level-3 'mostly efficient' anchor. | 4 / 5 |
Actionability | Core paths are copy-paste ready — the `createClient().use(...)` chain with `transactionConfig: { version: 1 }`, `NO_DNA=1 anchor build/test`, MCP install commands per host, `await createClient().use(surfpool())` — plus concrete API names (`walletSigner()`, `keychainSigner()`, `airdropSigner()`) and the `maxSupportedTransactionVersion: 1` read-side flag. However, several common task types (writing an Anchor/Pinocchio program, token creation, confidential transfers) get routing-only treatment with no inline example, matching 'mostly executable guidance; concrete code or commands with minor gaps' rather than the fully-covering level-5 anchor. | 4 / 5 |
Workflow Clarity | The five-step operating procedure (classify layer → pick blocks → implement with an explicit correctness checklist → add tests → deliverables with risk notes) is clearly sequenced, and the guardrails supply real validation checkpoints for risky operations ("Simulate before sending. Always run `simulateTransaction`", explicit approval before signing, devnet-by-default). It falls short of the level-5 anchor because error-recovery feedback loops are only implicit — failures route to common-errors.md rather than an explicit validate→fix→retry loop in the procedure itself. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview: all deep material (transactions-v1.md ~32KB, security.md ~34KB, testing.md ~21KB) lives in bundle files, and a dedicated "Progressive disclosure (read when needed)" index lists ~28 references with one-line scope descriptions. Every referenced path was verified to exist in the bundle, references are one level deep with no nested chains, and inline links are contextually placed. This matches 'clear overview with well-signaled one-level-deep references; easy navigation'. | 5 / 5 |
Total | 17 / 20 Passed |