Skip to content

docs: add OmniRoute admission lanes finding - #135

Open
branben wants to merge 1 commit into
modelcontextprotocol:mainfrom
branben:finding/admission-lanes
Open

docs: add OmniRoute admission lanes finding#135
branben wants to merge 1 commit into
modelcontextprotocol:mainfrom
branben:finding/admission-lanes

Conversation

@branben

@branben branben commented Sep 2, 2026

Copy link
Copy Markdown

Summary

Adds an experimental finding documenting OmniRoute's per-session virtual admission lanes as a production implementation of the runtime-layer complement to WG Topic #5 (tool dependencies).

What this demonstrates

  • Burst isolation: one session's tool burst doesn't 503 others
  • Fair scheduling via FairCostQueue round-robin
  • Idle eviction (60s TTL) for bounded memory
  • Backpressure with Retry-After headers
  • 109 tests passing in production 50k+ star repo

Relevant WG work

Security insight

PR #10110 (pre-auth resource exhaustion) reveals that fake credentials could shard admission lanes before auth. The WG should consider whether malicious declarations in frontmatter could exploit the same pattern.

Verification

git clone https://github.com/diegosouzapw/OmniRoute.git
cd OmniRoute
node --import tsx/esm --test tests/unit/admission-virtual-lanes-9654.test.ts
# Expected: all tests pass

Scope

Tests the dispatch layer, not the manifest layer. Does NOT implement the WG's proposed frontmatter field — implements the runtime gate that makes skill declarations safe under multi-user load.

Per-session virtual admission lanes as runtime-layer complement
to WG Topic modelcontextprotocol#5 (tool dependencies). Evidence from production
50k+ star repo with 109 tests passing.

References:
- WG Issue modelcontextprotocol#126 (Tool dependencies)
- OmniRoute PR #9654
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant