GitHub

### Feature: cancellation for host-owned external tools Host-owned external tool callbacks are now cancelled when their runtime request completes or their SDK session terminates. The cancellation primitive is idiomatic per SDK: .NET passes a request token to `AIFunction`, Node.js exposes `ToolInvocation.signal`, Go cancels `ToolInvocation.TraceContext`, Java cancels the returned `CompletableFuture`, Python cancels the handler task, and Rust drops the handler future. Go handlers that retain `TraceContext` for background work must derive a separate lifetime because the invocation context is cancelled when the request ends. ### Feature: declare application identity with client info Client options now accept optional client info (application name and version, integration name and version) across all six SDKs, exposed idiomatically per language (`clientInfo` in Node.js, `client_info` in Python and Rust, `ClientInfo` in Go and .NET, `setClientInfo` in Java). When set, the SDK forwards it on

In controlled offline evaluations, HydraFusion’s selective coding workflows matched or exceeded the evaluated Opus 5 baseline while reducing estimated workflow cost. Now available as a research preview in GitHub Copilot. The post Project HydraFusion: Frontier quality via multi-model orchestration appeared first on The GitHub Blog .

Learn how to run parallel agents in the GitHub Copilot app, and experience the moment it stops feeling scary and starts feeling powerful. The post GitHub Copilot app for Beginners: Run several agents at once appeared first on The GitHub Blog .

## 🌟 Release Highlights A focused patch release improving safe-output reliability and CI hygiene: `call_workflow` inputs no longer get dropped during samples replay, sample-coverage gaps are now surfaced as warnings, and AI credit budgets are validated more safely to close a CodeQL alert. ### 🐛 Bug Fixes & Improvements - **Fixed `call_workflow` samples replay dropping `workflow_name`** — ingestion previously stripped `workflow_name`/`inputs` for `call_workflow` safe outputs, causing the apply job to fail with "Workflow name is empty" ([#58113](https://github.com/github/gh-aw/pull/58113)). - **Added warnings for uncovered samples** — `--use-samples` compilation now warns when an enabled safe output has no `samples:` entries, so silent no-op replay runs are easier to catch ([#58112](https://github.com/github/gh-aw/pull/58112)). - **Hardened AI credit budget handling** — the max AI credits expression moved from inline shell interpolation to step `env:`, with strict integer validation befo

<!-- Release notes generated using configuration in .github/release.yml at v0.28.13 --> ## What's Changed ### Documentation * [docs] auth: list Google Vertex AI in README API-proxy provider list by @github-actions[bot] in https://github.com/github/gh-aw-firewall/pull/7957 * fix: patch vulnerable gRPC in GitHub CLI images by @github-actions[bot] in https://github.com/github/gh-aw-firewall/pull/7997 ### Other Changes * Add fail-closed Docker sbx egress verification by @lpcox in https://github.com/github/gh-aw-firewall/pull/7956 * Refactor shared Google provider adapter factories by @lpcox with @Copilot in https://github.com/github/gh-aw-firewall/pull/7954 * chore: upgrade gh-aw workflows by @lpcox in https://github.com/github/gh-aw-firewall/pull/7966 * ci: recompile Playwright smoke workflows by @lpcox in https://github.com/github/gh-aw-firewall/pull/7977 * Normalize parameterized model IDs for AI-credit pricing by @lpcox with @Copilot in https://github.com/github/gh-aw-firewall/pull/797

From loop engineering to harnesses, squads, and open weights, the GitHub Podcast breaks down the AI terms showing up in developer conversations. The post Decoding the new AI lingo: Loops, harnesses, squads, hill climbing&#8230; oh my! appeared first on The GitHub Blog .

Why shorter outputs can cost more, and how GitHub Copilot reduces wasted work across the complete coding task. The post How we make AI coding more cost efficient without sacrificing task quality appeared first on The GitHub Blog .

## What's Changed * Cover notification and star mutations in GitHub guard by @lpcox with @Copilot in https://github.com/github/gh-aw-mcpg/pull/12105 * [Repo Assist] refactor(config): extract shared non-empty string slice validator by @github-actions[bot] in https://github.com/github/gh-aw-mcpg/pull/12158 * [test] Add tests for server.createGuardFromConfig and requireGuardPolicyIfGuardEnabled by @github-actions[bot] in https://github.com/github/gh-aw-mcpg/pull/12110 * [log] Add debug logging to enclavegithub route matching by @github-actions[bot] in https://github.com/github/gh-aw-mcpg/pull/12129 * [test-improver] Improve tests for isValidRepoScope (guard_policy) by @github-actions[bot] in https://github.com/github/gh-aw-mcpg/pull/12130 * [Repo Assist] chore(server): bump MCPGatewaySpecVersion to 1.17.0 by @github-actions[bot] in https://github.com/github/gh-aw-mcpg/pull/12159 * [test] Add tests for proxy.enclaveRepositoryIsPublic and forwardEnclaveVisibilityLookup by @github-actions[bo

<!-- Release notes generated using configuration in .github/release.yml at v0.28.12 --> ## What's Changed ### Other Changes * fix: avoid self-matching KVM process probes by @lpcox in https://github.com/github/gh-aw-firewall/pull/7927 * fix(ci): match staged KVM processes by @lpcox in https://github.com/github/gh-aw-firewall/pull/7929 * Unexport LINUX_CAPABILITY_MAP from capability-filter.ts by @lpcox with @Copilot in https://github.com/github/gh-aw-firewall/pull/7928 * Add Playwright loopback smoke workflows by @lpcox in https://github.com/github/gh-aw-firewall/pull/7934 * Fix Playwright smoke tests across sandbox runtimes by @lpcox in https://github.com/github/gh-aw-firewall/pull/7938 * fix(cloud-hypervisor): limit device ACL lifetime by @lpcox in https://github.com/github/gh-aw-firewall/pull/7931 **Full Changelog**: https://github.com/github/gh-aw-firewall/compare/v0.28.11...v0.28.12 ## CLI Options ``` Usage: awf [options] [command] [args...] Network firewall for agentic workflows wi

<!-- Release notes generated using configuration in .github/release.yml at v0.28.11 --> ## What's Changed ### Documentation * [docs] auth: docs: reconcile Codex auto-model caveat between authentication-architecture and api-proxy-sidecar by @github-actions[bot] in https://github.com/github/gh-aw-firewall/pull/7860 ### Other Changes * chore: upgrade gh-aw prerelease to v0.87.9 by @lpcox in https://github.com/github/gh-aw-firewall/pull/7854 * api-proxy: honour explicit http:// scheme in target dialing instead of always dialing 443 by @lpcox with @Copilot in https://github.com/github/gh-aw-firewall/pull/7868 * docs: add B28 runner-doctor coverage for private-CA api-proxy TLS failures by @lpcox with @Copilot in https://github.com/github/gh-aw-firewall/pull/7871 * chore(deps): update safe patch dependencies by @lpcox with @Copilot in https://github.com/github/gh-aw-firewall/pull/7869 * fix: drop CAP_NET_ADMIN from Cloud Hypervisor VMM by @lpcox in https://github.com/github/gh-aw-firewall/pul

## 🌟 Release Highlights This release focuses on scheduling controls, trajectory-based agent evaluation, and hardened MCP Gateway/validation behavior. ### ✨ What's New - **`on.cooldown` workflow gating** ([#56998](https://github.com/github/gh-aw/pull/56998)) lets workflows enforce a minimum interval between runs, reducing noisy or redundant triggers. - **Typed `on.stop-after` field** ([#56983](https://github.com/github/gh-aw/pull/56983)) now supports GitHub Actions expressions, giving finer control over when scheduled/triggered workflows should stop running. - **Trajectory graders expanded** with new implementations for `policy-near-miss`, `exploitation-error`, `exploration-error`, `lempel-ziv-trajectory-complexity`, `skill-constraint-coverage`, and tool-output-consumption-rate ([#56996](https://github.com/github/gh-aw/pull/56996), [#57152](https://github.com/github/gh-aw/pull/57152), [#57087](https://github.com/github/gh-aw/pull/57087), [#56972](https://github.com/github/gh-aw/pull/569