Content
88%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, actionable gRPC hunting playbook with executable commands, a clear phased workflow, and strong false-positive validation discipline. Its main weakness is token efficiency and the absence of progressive disclosure: the full methodology is inlined in one file with no supporting bundle references.
Suggestions
Tighten the 'Crown Jewel Targets' preamble and the per-phase rationale paragraphs to the minimum needed to drive decisions; some explanatory context restates points already made in the Validation section.
Extract the Chain Table and/or the detailed proto-rebuild and gRPC-Web framing recipes into reference files under references/ (e.g. CHAINS.md, GRPC-WEB-FRAMING.md) and link to them one level deep, so SKILL.md reads as an overview.
Move the per-tool usage notes in the Tools section into a scripts/ or references/TOOLS.md, keeping only a one-line pointer for each tool inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient: the prose is largely non-obvious security judgment (status-code interpretation, why h2load cannot test rapid-reset, metadata-stripping mechanics) rather than concepts Claude already knows, but the ~270-line body has a few explanatory paragraphs (e.g. the 'Crown Jewel Targets' preamble) that could be trimmed. It sits clearly above the 'mostly efficient with some unnecessary explanation' anchor but is not fully lean. | 4 / 5 |
Actionability | Provides copy-paste-ready, executable commands throughout — nmap/openssl ALPN probes, grpcurl list/describe/call invocations, a complete protoscope-to-gRPC-Web framing recipe, and curl-based transcoded-route tests — covering the common cases with concrete tooling rather than pseudocode. | 5 / 5 |
Workflow Clarity | Clear seven-phase sequence with explicit validation checkpoints: the dedicated 'Validation — false-positive discipline' section defines feedback loops (status-code discrimination, re-run with grpcurl -v, authenticated-vs-unauthenticated state delta), and Phase 7 carries an authorization gate plus 'version-match instead of flooding' recovery guidance. Because validation is present for the destructive DoS and batch IDOR steps, the workflow-clarity cap of 3 does not apply. | 5 / 5 |
Progressive Disclosure | Well-organized with clearly signaled section headers (per-phase '##' headings, Chain Table, Tools, Validation) and easy in-file navigation, but no bundle files exist under references/scripts/assets and the entire ~270-line playbook is inlined in SKILL.md rather than split into one-level-deep reference files. This places it above 'content that should be separate is inline' but short of the 'overview pointing to detailed materials' ideal. | 4 / 5 |
Total | 18 / 20 Passed |