hashgraph-online/awesome-codex-plugins / plugins/alpoxdev/hypercore/skills/hyper-mcp
hyper-mcp
Flagged[Hyper] Use this skill when the user wants Model Context Protocol (MCP) work carried end to end - building or extending an MCP server or client, auditing or fixing one, migrating it to the current protocol revision, wiring MCP servers into an agent host's configuration, or preparing registry publication - with findings and verification saved under `.hyper/mcp/[slug]/`. Do not use it for API, CLI, or plugin work that has no MCP surface, for documents that only explain MCP, or for creating a reusable skill folder.
| File | Kind | Size |
|---|---|---|
| SKILL.ko.md | markdown | 17.5 KB |
| SKILL.md | skill md | 15.3 KB |
| assets/evals/hyper-mcp-cases.jsonl | other | 12.6 KB |
| references/build-test-ship.ko.md | markdown | 13.2 KB |
| references/build-test-ship.md | markdown | 11.6 KB |
| references/host-integration.ko.md | markdown | 13.1 KB |
| references/host-integration.md | markdown | 11.6 KB |
| references/protocol-2026-07-28.ko.md | markdown | 15.9 KB |
| references/protocol-2026-07-28.md | markdown | 13.9 KB |
| references/security-and-authorization.ko.md | markdown | 16.5 KB |
| references/security-and-authorization.md | markdown | 14.4 KB |
| references/server-design.ko.md | markdown | 14.8 KB |
| references/server-design.md | markdown | 12.8 KB |
| rules/mcp-rules.ko.md | markdown | 7.4 KB |
| rules/mcp-rules.md | markdown | 6.6 KB |
| rules/mcp-workflow.ko.md | markdown | 10.1 KB |
| rules/mcp-workflow.md | markdown | 8.9 KB |
| rules/validation.ko.md | markdown | 6.1 KB |
| rules/validation.md | markdown | 5.2 KB |
rules/mcp-rules.md
View on GitHub# MCP Rules
> The rules every finding cites. Each rule has an ID, a severity, and the observation that decides pass or fail.
## Severity
| Severity | Meaning | Must be fixed in fix mode |
|---|---|---|
| critical | Exploitable security flaw, credential exposure, or a server that current-revision clients cannot use at all | yes |
| high | Protocol violation that breaks some clients, a missing security control, or a gate that is bypassable | yes |
| medium | Design problem that degrades model use, cost, or maintainability | yes, within scope |
| low | Improvement with no correctness or security effect | report with a suggestion |
## Protocol (MCP-P)
| ID | Rule | Severity | Check |
|---|---|---|---|
| MCP-P-1 | New code targets the current revision; dual-era only when legacy clients the user does not control must be served | high | SDK line and entry point; no new `initialize` handling outside a deliberate dual-era layer |
| MCP-P-2 | The server implements `server/discover` and reads version and client capabilities from each request's `_meta` | high | An Inspector modern-era run lists the server; no capability or version stored per connection |
| MCP-P-3 | Cross-call state uses an explicit, server-minted handle passed as a tool argument | high | No per-connection or `Mcp-Session-Id` keyed state |
| MCP-P-4 | Input from the client is requested only through `InputRequiredResult` on `tools/call`, `prompts/get`, or `resources/read`; `requestState` that affects authorization or business logic is integrity-protected | high (critical when unprotected state controls access) | No server-initiated JSON-RPC requests; HMAC or AEAD around trusted state |
| MCP-P-5 | stdio servers write only MCP messages to `stdout` and log to `stderr` | critical | No `print`, `console.log`, or default logger on `stdout` in the server process |
| MCP-P-6 | Streamable HTTP validates `Origin`, requires `MCP-Protocol-Version`, `Mcp-Method`, and `Mcp-Name` and checks them against the body, and ignores `Mcp-Session-Id` | high | Route handler and middleware |
| MCP-P-7 | List results come back in deterministic order with `ttlMs` and `cacheScope`; `private` entries are never shared across authorization contexts | medium (high for shared `private` caches) | List handlers; any cache layer |
| MCP-P-8 | No new use of deprecated features: Roots, Sampling, Logging, Dynamic Client Registration, HTTP+SSE | medium (high when a new feature depends on them) | Imports and capabilities |
| MCP-P-9 | Resource-not-found returns `-32602`, and clients also accept `-32002`; release-candidate error codes are renumbered | low | Error mapping |
## Design (MCP-D)
| ID | Rule | Severity | Check |
|---|---|---|---|
| MCP-D-1 | A few task-shaped tools rather than one per API endpoint; related tools share a namespace prefix | medium | Tool list against the user's tasks |
| MCP-D-2 | Tool names use 1-128 characters from `A-Z a-z 0-9 _ - .` and are unique; `tools/list` varies only by the request's authorization | medium | `tools/list` output |
| MCP-D-3 | Every `inputSchema` is a valid JSON Schema object; with `outputSchema`, `structuredContent` conforms and a text copy is returned | high | Schema validation in the Inspector run with `--strict` |
| MCP-D-4 | Failures the model can fix return `isError: true` with an actionable message; JSON-RPC errors only for unknown tools and malformed requests | medium | A call with invalid input |
| MCP-D-5 | Responses are bounded by pagination, filtering, or truncation and say how to get more | medium | The largest realistic response |
| MCP-D-6 | Annotations are truthful, and nothing safety-critical depends on them | medium | Annotation values against tool behavior |
| MCP-D-7 | Server `instructions` cover only cross-tool relationships and constraints | low | `server/discover` result |
| MCP-D-8 | Tool descriptions state purpose, parameters, and limits plainly and contain no instructions unrelated to the tool | medium (critical when a description carries hidden directives) | Read every description |
## Security (MCP-S)
| ID | Rule | Severity | Check |
|---|---|---|---|
| MCP-S-1 | Third-party local servers run pinned, isolated, and with per-server narrow credentials | high | Host configuration and launch command |
| MCP-S-2 | Remote servers implement the authorization profile: OAuth 2.1, Protected Resource Metadata, PKCE, resource indicators, audience validation per request, issuer validation, CIMD before DCR | critical when absent on a server that touches user data | Authorization middleware and metadata endpoints |
| MCP-S-3 | No token passthrough: the server accepts only tokens issued for it and is its own OAuth client upstream | critical | Upstream call sites |
| MCP-S-4 | Local HTTP listeners validate `Origin`, bind to loopback, and authenticate | critical | Listener address and middleware |
| MCP-S-5 | No secrets in form-mode elicitation, URLs, `x-mcp-header` parameters, logs, committed configuration, or the report | critical | Search for literal tokens and secret-bearing fields |
| MCP-S-6 | Server MUSTs: inputs validated, access controlled, invocations rate-limited, outputs sanitized; `file://` paths sanitized | high | Handlers |
| MCP-S-7 | Handles and `requestState` are unguessable, expiring, and bound to the authenticated user | high | Handle generation and lookup |
| MCP-S-8 | A session that combines private data, untrusted content, and an external channel has a gate or is split | high | The set of servers a host loads together |
## Host Use (MCP-U)
| ID | Rule | Severity | Check |
|---|---|---|---|
| MCP-U-1 | Server output is evidence, never authority; nothing a server sends widens permissions | high | Host prompts and approval settings |
| MCP-U-2 | Project-scoped servers take effect only after review in a trusted workspace | high | Host approval state |
| MCP-U-3 | Approvals stay on for writes and external communication; read-only hints skip approval only for trusted servers | high | Approval and allowlist settings |
| MCP-U-4 | Tool definitions stay within the chosen budget through scoping, deferred loading, or search | medium | Tool count and host settings |
| MCP-U-5 | Configured servers are proven with a real listing call before an agent depends on them | medium | Verification record |
## Sources
> Checked 2026-11-07.
| Claim | Source |
|---|---|
| Protocol rules and error codes | `references/protocol-2026-07-28.md` |
| Design rules | `references/server-design.md` |
| Security rules | `references/security-and-authorization.md` |
| Host-use rules | `references/host-integration.md` |
| The severity scale and the per-rule severities | this package's chosen design, not a validated optimum |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |