Skip to content

feat: ship Buzzr public sports engine and MCP ecosystem#10

Merged
sarveshsea merged 91 commits into
mainfrom
feat/public-mcp-toolkit
Jul 16, 2026
Merged

feat: ship Buzzr public sports engine and MCP ecosystem#10
sarveshsea merged 91 commits into
mainfrom
feat/public-mcp-toolkit

Conversation

@sarveshsea

@sarveshsea sarveshsea commented Jul 16, 2026

Copy link
Copy Markdown
Member

Objective

Ship Buzzr as a complete public sports-settlement and agent toolkit: hardened engine behavior, an 11-tool stdio MCP server, a repository-owned Codex skill, complete package and API documentation, exact release automation, official registry metadata, and measurable adoption baselines.

What ships

  • correct, fail-safe DFS settlement behavior with versioned conformance vectors and current PrizePicks push, tie, and DNP semantics
  • deterministic fixed-table payout selection, pending results for unsupported zero-loss rows, and validated verification metadata
  • @buzzr/mcp with 11 tools, bounded input, schema rejection, correlated concurrency, malformed-input recovery, stdout purity, and clean shutdown
  • clean npx -y @buzzr/mcp startup plus real stdio client and cross-platform packed-artifact tests
  • repository skill at skills/buzzr-sports-engine with validated examples, discovery, troubleshooting, and isolated install proof
  • package README coverage for all ten workspaces plus generated all-package TypeDoc API docs
  • official MCP Registry metadata for io.github.Buzzr-app/dfs-engine
  • adoption capture for all npm packages, GitHub traffic, issues, referrers, popular paths, and official registry presence

Release train

Five changed public packages are authorized by release-manifest.json:

  • @buzzr/dfs-engine@5.1.0
  • @buzzr/dfs-engine-test-vectors@5.1.0
  • @buzzr/mcp@5.1.0
  • @buzzr/dfs-cli@5.0.1
  • @buzzr/dfs-testkit@5.0.1

The other five workspaces remain at 5.0.0. Internal dependency pins and all ten versions are release-contract checked.

Verification

  • full npm run verify
  • 394 engine tests and 95 MCP tests
  • engine coverage: 91.8% statements, 86.3% branches
  • ten clean-installed packed packages with ESM, CJS, type, bin, provider, testkit, React, and vector replay proof
  • real packed MCP initialization and 11-tool discovery
  • repository skill validation with the pinned official validator and isolated installer
  • 42 documentation files, 107 local references, and 69 external URLs checked
  • package-size and export smoke checks
  • release workflow and supply-chain regression contracts
  • performance benchmark
  • tracked-file secret scan
  • npm audit: 0 vulnerabilities

Release safety

  • build, test, staging, and packing run without OIDC privileges
  • exactly five SHA-512-bound tarballs are transferred through SHA-pinned artifact actions
  • npm OIDC job publishes only literal reviewed tarballs with lifecycle scripts disabled and provenance enabled
  • all five registry versions are preflighted before mutation; conflicts fail closed
  • exact partial releases can resume safely after main advances, but new releases still require the reviewed commit to be current main
  • npm proof, MCP Registry OIDC, tag creation, and GitHub release are separated and bound to the reviewed commit and hashes
  • exact tag creation and release notes are deterministic and rerunnable

Independent review

Final code and security reviews both returned GO on commit 26d31816aabc5d8ab9da62d8a1124ed72a354bce with no residual findings.

@sarveshsea sarveshsea changed the title feat: harden Buzzr public MCP toolkit feat: ship Buzzr public sports engine and MCP ecosystem Jul 16, 2026
@sarveshsea
sarveshsea marked this pull request as ready for review July 16, 2026 16:46
@sarveshsea
sarveshsea merged commit a4d4a2f into main Jul 16, 2026
4 checks passed
@sarveshsea
sarveshsea deleted the feat/public-mcp-toolkit branch July 16, 2026 16:50
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