v0.82.2

🌟 Release Highlights

This release focuses on ARC/DinD (Actions Runner Controller + Docker-in-Docker) reliability, expanded skills: frontmatter capabilities, and a wave of quality improvements to ESLint rules, safe-outputs, and the Windows CI matrix.

✨ What's New

• ARC/DinD topology is now fully supported. Three coordinated fixes (#42815, #42855, #42906) redirect runtime-critical paths to daemon-visible locations (${RUNNER_TEMP}/gh-aw), redirect DOTNET_INSTALL_DIR and GOPATH, and pre-create required mount directories — so compiled workflows run reliably on ARC/DinD runners without path-visibility failures. Requires AWF firewall ≥ v0.27.22.

• Per-skill GitHub auth in skills: frontmatter. Each skill entry can now carry its own github-token or github-app credential, enabling multi-org or multi-repo skill installs with distinct authentication. (#42594)

• Runtime reviewers, team-reviewers, and assignees in create-pull-request. Reviewer and assignee fields now accept template expressions resolved at runtime — routing the created PR to the triggering actor or any dynamically computed value. (#42621)

• PR Sous Chef auto-dismisses stale bot reviews. After an agent addresses feedback, pr-sous-chef can now dismiss stale CHANGES_REQUESTED reviews from github-actions[bot], letting PRs proceed without manual cleanup. (#42626)

• reasoning_depth A/B experiment in plan workflow. A three-variant experiment (shallow, baseline, deep) measures whether lighter or deeper prompts improve sub-issue quality and cost. (#42955)

🐛 Bug Fixes & Improvements

• Safe-outputs pre-flight full-branch validation — the allowed_files check now validates the full commit history reachable from a pushed ref, preventing false-safe pushes when older commits touched .github/workflows/. (#42585)
• Skill install failures surface as agent failure context — when gh skill install fails, the failure is detected and attached to the agent's failure context for clearer diagnostics. (#42642)
• BYOK provider inheritance for sub-agents — fixes sub-agents not inheriting the BYOK provider from the headless sidecar environment. (#42613)
• Windows cmd/direct quoting fix — resolves a .NET vs cmd.exe quote escaping mismatch causing cmd/direct scenario failures in the Windows CI matrix. (#42951)
• Workspace wipe prevention during local-actions checkout — generated checkout steps now include clean: false so .git state is preserved for subsequent local-action post-steps. (#42611)
• ESLint rule improvements — no-unsafe-promise-catch-error-property brought to parity with its sibling rule, null-safe companion requirement for typeof err === "object" catch guards, and new require-async-entrypoint-catch rule. (#42953, #42568, #42935)
• timesleepnocontext linter boundary fix — stops incorrectly crossing non-Go/defer FuncLit boundaries when attributing time.Sleep calls. (#42946)

📚 Documentation

• skills: frontmatter reference — new documentation covering pinned refs, per-skill auth, and a real-world complex example. (#42747)
• Safe outputs reference slimmed down — the safe outputs reference is now significantly shorter and easier to scan. (#42814)
• app: → github-app: migration completed — gh aw fix now covers top-level app: blocks in addition to nested occurrences. (#42794)

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by 🚀 Release · 42.4 AIC · ⊞ 7.7K

---

What's Changed

• fix(pr-code-quality-reviewer): prevent timeout by launching grumpy-coder early by @pelikhan with @Copilot in #42583
• doc-healer/doc-updater: classify site-build/UI issues as non-content, skip them from docs-PR path by @pelikhan with @Copilot in #42590
• Align sandbox integration tests with rootless AWF command default by @pelikhan with @Copilot in #42606
• Add tone_variant A/B campaign to breaking-change-checker workflow by @pelikhan with @Copilot in #42597
• Promote errortypeassertion and execcommandwithoutcontext to blocking CI gate by @pelikhan with @Copilot in #42580
• fix(eslint-factory): require-parseInt-radix false negative on radix 0/undefined by @pelikhan with @Copilot in #42581
• fix: detect "no model endpoints available" as http_400_response_error by @pelikhan with @Copilot in #42593
• eslint-factory: require null-safe companion for typeof err === "object" catch guards by @pelikhan with @Copilot in #42568
• fix: wire BYOK provider into headless sidecar env so sub-agents inherit it by @pelikhan with @Copilot in #42613
• docs: add opt-in, participant-targeted recruitment banner to Starlight docs by @pelikhan with @Copilot in #42595
• Prevent workspace wipe during local-actions checkout in generated workflows by @pelikhan with @Copilot in #42611
• Refactor schema compilation to use a single shared helper by @pelikhan with @Copilot in #42614
• build(deps-dev): bump starlight-github-alerts from 0.3.0 to 0.4.0 in /docs by @dependabot[bot] in #42462
• build(deps): bump mermaid from 11.15.0 to 11.16.0 in /docs by @dependabot[bot] in #42459
• build(deps-dev): bump prettier from 3.8.4 to 3.9.4 in /actions/setup/js by @dependabot[bot] in #42458
• build(deps): bump starlight-blog from 0.26.1 to 0.27.0 in /docs by @dependabot[bot] in #42455
• build(deps-dev): bump vite from 8.0.16 to 8.1.0 in /actions/setup/js by @dependabot[bot] in #42456
• [community] Update community contributions in README by @github-actions[bot] in #42616
• Support per-skill GitHub auth in frontmatter skills by @pelikhan with @Copilot in #42594
• PR Sous Chef: dismiss stale github-actions[bot] blocking reviews after feedback is addressed by @pelikhan with @Copilot in #42626
• docs: move trigger decision matrix to triggers.md, add coverage/digest fallback guidance by @pelikhan with @Copilot in #42618
• Extend windows-cli-integration scenario matrix: add cmd/direct and path-launch for non-default shells by @pelikhan with @Copilot in #42622
• Add clean: false to agentics-maintenance.yml checkout steps by @pelikhan with @Copilot in #42623
• compact Test Quality Sentinel report template (~480 chars saved) by @pelikhan with @Copilot in #42633
• Add improvement issue creation to daily GEO optimizer by @pelikhan with @Copilot in #42635
• fix(pr-sous-chef): use sonnet alias for pr-processor sub-agent to resolve HTTP 400 by @pelikhan with @Copilot in #42634
• feat(uk-ai-resilience): create up to 2 labeled issues for top-priority findings by @pelikhan with @Copilot in #42636
• [log] Add debug logging to 5 pkg/ files by @github-actions[bot] in #42640
• refactor: extract helpers to bring 3 functions under the 60-line largefunc limit by @pelikhan with @Copilot in #42624
• refactor: skip gh CLI upgrade when already at minimum version for skill support by @pelikhan with @Copilot in #42639
• [code-simplifier] simplify: reduce duplication in discussion comment handlers by @github-actions[bot] in #42641
• safe-outputs: pre-flight workflow scope check + full-branch allowed_files validation by @pelikhan with @Copilot in #42585
• Improve time-delta validation diagnostics to satisfy error-message lint quality by @pelikhan with @Copilot in #42651
• Cover wasm-only files in custom li...

v0.82.1

🌟 Release Highlights

v0.82.1 focuses on a major dashboard overhaul, new workflow authoring primitives, expanded static analysis coverage, and reliability fixes across the compiler and runtime.

⚠️ Behavioral Change: sandbox.agent.sudo Defaults to false

The global default for sandbox.agent.sudo has changed from true to false (network isolation / rootless mode). Workflows that relied on elevated sandbox access must now explicitly set sandbox.agent.sudo: true. This aligns the default with the recommended security posture.

✨ What's New

• Frontmatter skills support — Workflows can now declare SHA-pinned skills directly in frontmatter. At activation time, gh aw automatically installs them via gh skill install and wires them into the engine. Learn more (#42426, #42543)

• Model policy controls — New models.allowed and models.disallowed frontmatter fields let you restrict which AI models a workflow can use, with environment-variable overrides for flexibility. (#41824)

• ARC/DinD rootless topology — runner.topology: arc-dind is now a first-class frontmatter option, enabling GitHub ARC runners with Docker-in-Docker rootless execution. (#42371)

• branding field in aw.yml — Package manifests can now include a GitHub Actions branding block without failing schema validation. (#42454)

• Per-model Copilot SDK wireApi — Updated for Copilot SDK v1.0.4's multi-provider BYOK registry, where transport behavior can vary by model. (#42497)

🖥️ Dashboard Overhaul

The agentic workflows dashboard received extensive polish this release:

• TypeScript migration — Backend modules migrated from ESM .mjs to strict TypeScript, eliminating all any types and narrowing unknown (#42321, #42353)
• Bundled into single artifacts — Alpine.js and Node.js code are now bundled into self-contained app.js — no external CDN dependencies (#42300, #42352)
• Skeleton loaders & spinners — Loading states are now visually communicated while runs and definitions fetch (#42345)
• Cross-session log caching — Downloaded logs are cached in a shared user-level directory, reducing redundant fetches (#42349)
• Compact single-line rows — Definitions and Runs lists now render in tighter, more scannable single-line format (#42344)
• Missing gh binary detection — Dashboard shows a targeted install prompt when the gh CLI is not found (#42303)
• Fixed hang on inherited handles — Detached spawn prevents gh aw from hanging when dashboard subprocess inherits stdio handles (#42311)

🐛 Bug Fixes & Improvements

• Fixed concurrent schedule overlap — every N minutes cron expressions are now scattered across different offsets, preventing all same-interval workflows from firing simultaneously (#42524)
• Hardened API fallbacks — Unauthenticated api.github.com calls now carry a 30-second timeout and request context, eliminating potential goroutine hangs during compilation (#42521)
• Fixed AWF access.log path — Firewall log analysis now correctly finds access.log in the squid-logs/ subdirectory (#42513)
• Codex retry loop fix — Codex harness no longer retries on TPM exhaustion or unfinished-goal errors, avoiding run quota drain (#42420)
• Better engine typo detection — Invalid engine: values now surface as the primary error rather than being masked by downstream schema failures (#42235)
• Auto-derive checkout path: — Compiler now warns and auto-derives the path: for cross-repo checkout entries where it was missing (#42257)

🔍 New Static Analysis Rules

• sprintfint Go linter — Flags fmt.Sprintf("%d", x) where x is int, suggesting strconv.Itoa for clarity (#42538)
• errortypeassertion Go analyzer — Catches direct error-to-concrete type assertions and steers code toward errors.As (#42323)
• httpnoctx enhancement — Now detects the http.NewRequest(...) + client.Do(req) path, not just helper functions (#42536)
• ESLint no-core-setoutput-non-string — Flags non-string values passed to core.setOutput, preventing type coercion bugs in Actions JavaScript (#42453)

📚 Documentation

• Aligned gh aw CLI help text with docs for version, deploy, strict, and secrets flags (#42496)
• Streamlined MemoryOps guide and troubleshooting common issues pages (#42533, #42308)
• Added sandbox.agent.sudo glossary entry documenting the AWF privilege model (#42466)

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by 🚀 Release · 64.3 AIC · ⊞ 7.7K

---

What's Changed

• Use Sonnet consistently in pr-sous-chef by @pelikhan with @Copilot in #42292
• feat: add audit logging to comment_release_prs job by @pelikhan with @Copilot in #42306
• [docs] docs: unbloat troubleshooting common issues by @github-actions[bot] in #42308
• Skip signed/jsweep/* branches in PR Description Updater by @pelikhan with @Copilot in #42296
• Refactor forecast command into focused modules and isolate render helper tests by @pelikhan with @Copilot in #42274
• fix(dashboard): button group styling and Alpine.js FOUC prevention by @pelikhan with @Copilot in #42309
• fix(aw-dashboard): use detached spawn to prevent gh-aw hang on inherited handles by @pelikhan in #42311
• Add maintenance view to the agentic workflows dashboard by @pelikhan with @Copilot in #42297
• Scale MCP logs timeout for larger fetch windows by @pelikhan with @Copilot in #42295
• feat(dashboard): detect missing gh binary and show targeted install prompt by @pelikhan with @Copilot in #42303
• fix(contribution-check): eliminate E002 safe-output collisions and retry loops by @pelikhan with @Copilot in #42316
• docs(aw): add compliance and design-governance patterns to authoring guidance by @pelikhan with @Copilot in #42315
• Bundle Alpine.js into the dashboard extension by @pelikhan with @Copilot in #42300
• Simplify gh aw status table output by @pelikhan with @Copilot in #42312
• Prioritize invalid engine typos over downstream schema/import failures by @pelikhan with @Copilot in #42235
• Honor global mention allowlists in add-comment sanitization by @pelikhan with @Copilot in #42313
• Normalize report-style prompts in Priority 1 workflows (h3+ headers + <details> guidance) by @pelikhan with @Copilot in #42324
• Migrate dashboard backend modules from ESM .mjs to strict TypeScript by @pelikhan with @Copilot in #42321
• Allow create_pull_request bundle transport to fall back when branch pinning is unavailable by @pelikhan with @Copilot in #42318
• dashboard: memoize binary detection and robust JSON parsing by @pelikhan in #42334
• fix(dashboard): alert box spacing and scrollbar polish by @pelikhan with @Copilot in #42335
• dashboard: add left margin to Counter badges by @pelikhan with @Copilot in #42337
• dashboard: structured logging, fix cwd/spawn bugs, add build:ts + AGENTS.md by @pelikhan in #42338
• Extract shared SafeOutputAllowBlockConfig across safe-output handlers by @pelikhan with @Copilot in #42322
• dashboard: compact single-line rows for Definitions and Ru...

v0.82.0

🌟 Release Highlights

This release brings a new Copilot Canvas extension for agentic workflow operations, significant dashboard improvements, experimental LSP frontmatter support, and an expanded ESLint Factory with new safety rules.

✨ What's New

• 🎨 Copilot Canvas Extension — A new project-scoped GitHub Copilot canvas extension (agentic-workflows-dashboard) lets you browse workflow definitions and runs, review safe markdown step summaries, and dispatch workflows directly from the Copilot chat panel. (#42137)

• 📊 Dashboard: Experiments Tab — The dashboard canvas now surfaces A/B experiment branch data via a dedicated Experiments tab backed by gh aw experiments list. (#42213)

• 📈 Dashboard: Usage Reporting & Continuation-Aware Log Windows — Run details now include token usage reporting and context-aware log windows that track continuation state. (#42226)

• 🔬 Dashboard: Audit Data with Progressive Disclosure — Run detail panels now surface structured audit data inline, with progressive disclosure to keep the view clean. (#42269)

• 🛠️ LSP Frontmatter Support (Experimental) — Workflows can now declare lsp: frontmatter configuration so Copilot CLI automatically receives generated lspServers config and required server dependencies during agent setup. Non-Copilot engines fail fast with a clear error. (#41777)

• 📏 ESLint Factory — New Safety Rules — Three new rules land in the ESLint Factory: no-unsafe-catch-error-property (flags unsafe .message, .stack, .status, .cause, .name access), no-unsafe-promise-catch-error-property (same for promise .catch() callbacks), and require-parseInt-radix (enforces explicit radix argument). (#42057, #42221, #41961)

• 🔁 Templatable report-failure-as-issue — Reusable workflows can now parameterize the safe-outputs.report-failure-as-issue field without post-compile lock-file edits, enabling flexible failure reporting across shared workflow definitions. (#41821)

• 📁 Repo-Local Instruction Overlays — Installed gh-aw agents now support repo-local instruction overlay files, allowing per-repository customization of agent behavior without forking upstream workflows. (#42053)

🐛 Bug Fixes & Improvements

• Safe-outputs set_issue_type — Fixed incorrect GraphQL query and field name in the set_issue_type intent path. (#42232)
• MCP Gateway config — Fixed invalid JSON escape sequences in GitHub remote MCP gateway config generation. (#41864)
• Setup cleanup — gh aw setup now properly cleans up root-owned /tmp/awf-*-chroot-home directories. (#41852)
• Copilot HTTP 400 failures — HTTP 400 responses from Copilot are now surfaced in workflow conclusions rather than silently discarded. (#42228)
• dispatch-repository key — Enforced canonical dispatch-repository safe-output key and deprecated the underscore alias dispatch_repository. (#42150)
• Audit job steps — Structured audit output now correctly populates individual job step data. (#42222)
• Enterprise host context — Enterprise host configuration is now properly propagated into the curated DIFC/CLI proxy environment. (#41912)

📚 Documentation

• Corrected gh aw mcp add argument order in the setup guide. (#42272)
• Added dedup checklist, PR escalation guidance, and linked-reference patterns to developer docs. (#42111)
• Added Agent Finder discovery guidance with direct REST API examples for skill selection. (#42078)

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by 🚀 Release · 35.4 AIC · ⊞ 7.7K

---

What's Changed

• fix: apply_samples derivePrHeadRef uses target-repo config for siderepo PR lookups by @dsyme with @Copilot in #41295
• Enable templatable report-failure-as-issue and shared-workflow propagation by @pelikhan with @Copilot in #41821
• Add nolint suppression parity for 5 remaining gh-aw analyzers by @pelikhan with @Copilot in #41849
• fix(fmterrorfnoverbs): treat %% as escaped literal, not a real verb by @pelikhan with @Copilot in #41850
• [instructions] Sync instruction files with release v0.81.6 by @github-actions[bot] in #41871
• [spec-extractor] Update package specifications for envutil, fileutil, gitutil, logger by @github-actions[bot] in #41873
• [docs] Consolidate developer specifications: v9.19 maintenance tone scan by @github-actions[bot] in #41887
• [docs] Update editor preview screenshots – 2026-06-27 by @github-actions[bot] in #41868
• Fix invalid JSON escapes in GitHub remote MCP gateway config generation by @pelikhan with @Copilot in #41864
• Bump Claude Code to 2.1.195 and Codex to 0.142.3 by @pelikhan with @Copilot in #41865
• fix(log_parser_bootstrap): guard Claude guardrail with safeOutputEntriesCount escape hatch by @pelikhan with @Copilot in #41886
• fix(setup): clean up root-owned /tmp/awf-*-chroot-home directories by @pelikhan with @Copilot in #41852
• Gate CI coverage uploads behind code-quality opt-in variable by @pelikhan with @Copilot in #41922
• Add Copilot-only lsp frontmatter support with schema, codegen, engine guardrails, and LSP instructions (experimental) by @pelikhan with @Copilot in #41777
• Strengthen anomaly detector edge coverage and remove order-dependent test behavior by @pelikhan with @Copilot in #41916
• [linter-miner] linter: add httpstatuscode analyzer — flag magic HTTP status code literals by @github-actions[bot] in #41914
• [formal-spec] Add OTEL observability formal compatibility test suite (v0.4.0 L1/L2) by @pelikhan with @Copilot in #41906
• Add run-again footer hints for slash and label command triggers by @pelikhan with @Copilot in #41920
• refactor: extract shared CLI engine execution logic; make Pi backend resolution configurable by @pelikhan with @Copilot in #41921
• fix(changeset): restrict safe-outputs allowed-files to .changeset/*.md by @pelikhan with @Copilot in #41937
• [caveman] Optimize instruction verbosity — memory.md, mcp-clis.md (2026-06-27) by @github-actions[bot] in #41943
• safeoutputs: disambiguate submit_pull_request_review.event from action by @pelikhan with @Copilot in #41948
• Establish ESLint Linter Factory for actions/setup/js and add dedicated ESLint daily workflows by @pelikhan with @Copilot in #41938
• pr-sous-chef: auto-resolve outdated addressed review threads via thread-resolver sub-agent by @pelikhan with @Copilot in #41959
• feat(ambient-context): trim prompt size in daily-code-metrics, test-quality-sentinel, smoke-copilot-aoai-entra by @pelikhan with @Copilot in #41960
• eslint-factory: add hasSuggestions, enriched diagnostic message, and try/catch suggest fix by @pelikhan with @Copilot in #41966
• [eslint-miner] feat(eslint): add require-parseInt-radix rule — ESLint Miner 2026-06-27 by @github-actions[bot] in #41961
• Clarify LSP capabilities and prompt patterns in lsp.md for authoring efficiency by @pelikhan with @Copilot in #41977
• Add ready_for_review workflow for Impeccable (needex) skills-based PR review by @pelikhan with @Copilot in #41978
• Moving ESLint factory to new location by @pelikhan with @Copilot in #41975
• eslint-factory: fix require-json-parse-try-catch to catch computed JSON["parse"] access by @pelikhan with @Copilot in #42000
• Bump gh-aw-firewall to v0.27.12 and gh-aw-mcpg to v0.3.31 by @lpcox with @Copilot in...

v0.81.6

🌟 Release Highlights

This release focuses on stability and observability — restoring broken fleet-wide token tracking, hardening CI quality gates, and adding release traceability for merged PRs.

✨ What's New

• Release PR notifications — A new post-agent release job now automatically comments on every PR included in each release, making it easy to trace which release shipped your changes (#41834).
• Loop engineering playbook — .github/aw/loop.md codifies patterns from autoloop, goal, and crane into a unified reference for building robust agentic loops (#41833).
• Benchstat regression gate — The CI bench job now compares against a stored baseline using benchstat, preventing silent performance regressions from merging (#41813).

🐛 Bug Fixes & Improvements

• Restored fleet-wide token usage collection — TokenUsage had been reporting 0 across the entire fleet since June 20 due to two co-conspirating bugs in the conclusion job. Both are fixed, restoring accurate AI credit tracking (#41823).
• Security: pinned govulncheck to go.mod version — CI vulnerability scans now use the exact version declared in go.mod and produce reproducible SARIF reports. A local make vuln-sarif target is also available (#41815).
• Hardened BYOK Ollama startup — The daily BYOK Ollama test now gates on explicit model and API readiness before proceeding, eliminating pre-agent failures from incomplete startup (#41838).
• Stabilized Go Logger Enhancement compile — Fixed exit code 126 caused by unsafe shell glob expansion in the workflow tool allowlist (#41840).
• Enforced safe output completion in quality workflow — The daily compiler quality workflow now requires an explicit safe output call, preventing silent no-op runs (#41841).

🔧 Internal

• Refactored 5 extreme function-length hotspots (145–650 lines) in pkg/workflow and pkg/cli into focused, testable units (#41800).
• SortedKeys sweep and deduplication logic consolidation across the codebase (#41829).

Generated by 🚀 Release · 32.9 AIC · ⊞ 8.3K

---

What's Changed

• refactor: split 5 extreme function-length hotspots in pkg/workflow and pkg/cli by @pelikhan with @Copilot in #41800
• fix(security): pin govulncheck to go.mod version in CI, add local SARIF target by @pelikhan with @Copilot in #41815
• Add benchstat regression gate to CI bench job by @pelikhan with @Copilot in #41813
• fix(token-usage): restore fleet-wide TokenUsage collection via non-empty copy guard and correct priority order by @pelikhan with @Copilot in #41823
• Add post-agent release job to comment on PRs included in each release by @pelikhan with @Copilot in #41834
• [community] Update community contributions in README by @github-actions[bot] in #41836
• Require explicit safe output completion in daily compiler quality workflow by @pelikhan with @Copilot in #41841
• Harden Daily BYOK Ollama startup path with explicit model/API readiness gating by @pelikhan with @Copilot in #41838
• [aw] Stabilize Go Logger Enhancement compile invocation by @pelikhan with @Copilot in #41840
• refactor: SortedKeys sweep, deduplicate dedup logic, redistribute misplaced helpers by @pelikhan with @Copilot in #41829
• Add .github/aw/loop.md playbook synthesizing loop-engineering patterns from autoloop/goal/crane by @pelikhan with @Copilot in #41833

Full Changelog: v0.81.5...v0.81.6

v0.81.5

🌟 Release Highlights

This release focuses on org-mode maturity, performance optimizations, and sandbox hardening — making gh aw update/upgrade --org more robust and production-ready while delivering meaningful speed improvements across the compiler and validator.

✨ What's New

• Action-pin mapping in aw.json (#41579) — Define action-pin overrides directly in your aw.json configuration, giving you centralized control over pinned action versions.

• manifest-version now optional in aw.yml (#41687) — Reduces boilerplate in new workflow files; the field is inferred when omitted.

• Non-TTY fallback for gh aw add wizard (#41717) — The interactive add wizard now gracefully falls back to text prompts in non-interactive environments (CI, scripts), unblocking automation pipelines.

• Org-mode improvements for update/upgrade --org (#41617, #41627, #41719) — Workflow-targeted updates, repo prefiltering, version-tag display, unified repo discovery, and per-repo confirmation prompts (with --yes for CI). The --org flag is now significantly more useful for managing fleets of repositories.

• Sandbox hardening at 50% baseline (#41786) — Half of all eligible agentic workflows now run with sandbox.agent.sudo: false, reducing the blast radius of runaway agent steps.

• Daily detection analysis report workflow (#41802) — New built-in workflow that generates automated detection analysis reports.

⚡ Performance

• Lazy-loaded embedded JSON datasets (#41587) — Embedded compile-time JSON is no longer loaded at startup, reducing memory footprint for CLI commands that do not compile.
• Lazy-loaded GitHub toolset permissions (#41755) — Loaded via sync.Once so permission data is only read when needed.
• Cached regexp in applySanitizePattern (#41762) — Eliminates repeated regex compilation on hot code paths.
• Faster ParseWorkflow (#41772) — Model-alias cycle detection overhead reduced.

🐛 Bug Fixes & Improvements

• Secret double-escape fix (#41801) — Custom MCP server env/header secrets are no longer double-escaped in generated lock files.
• Agent-supplied branch accepted in push_to_pull_request_branch (#41654) — Fixes a regression where agent-provided branch names were rejected.
• MODEL_NOT_SUPPORTED detection extended (#41792) — The pattern now catches 404 "Model not found" responses, improving resilience for unsupported model errors.
• Bundle manifest path resolution (#41790) — .github/ paths are now resolved as repo-root-relative in nested bundle manifests.
• MCP post-completion relaunch is non-fatal (#41713) — A failed MCP server relaunch after job completion no longer fails the workflow.
• PR Sous Chef cooldown (#41759) — Prevents back-to-back comments by enforcing a 30-minute cooldown.
• Harness exits cleanly (#41675) — Exits with code 0 when expected safe-outputs were already produced, even when subsequent steps encounter permission-denied errors.
• GH_AW_POLICY_STRICT enforced at runtime (#41682) — Non-strict compiled workflows now properly respect the strict policy flag at runtime.

📚 Documentation

• Blog: Custom Linters, Sergo, Linter Miner & LintMonster (#41663) — A new blog post walking through the custom linter ecosystem. Read it →
• gh aw env governance guide (#41758) — New documentation covering defaults and scope inheritance for environment configuration.

Generated by 🚀 Release · 41.2 AIC · ⊞ 8.3K

---

What's Changed

• Add action-pin mapping support in aw.json by @pelikhan with @Copilot in #41579
• pkg/workflow: lazy-load embedded JSON datasets used only at compile-time paths by @pelikhan with @Copilot in #41587
• feat: add XML markers, deduplication, labels, and release links to org runner PRs/issues by @pelikhan with @Copilot in #41580
• fix: resource lifecycle and context propagation (lint-monster) by @pelikhan with @Copilot in #41589
• Replace lint-monster path literals with repository constants in CLI/workflow codepaths by @pelikhan with @Copilot in #41611
• [UX] Professionalize Architecture Guardian safe-output status messages by @pelikhan with @Copilot in #41612
• [jsweep] Clean update_context_helpers.cjs by @github-actions[bot] in #41608
• fix: accept agent-supplied branch in push_to_pull_request_branch (fixes #41643) by @dsyme in #41654
• fix: replace deprecated gpt-5-mini with gpt-5.4-mini in daily-model-resolution workflow by @pelikhan with @Copilot in #41625
• Enable sandbox.agent.sudo: false on 20 agentic workflows by @pelikhan with @Copilot in #41628
• [docs] Update glossary - daily scan by @github-actions[bot] in #41657
• [spec-extractor] Update package specifications for agentdrain, cli, console, constants by @github-actions[bot] in #41653
• [code-scanning-fix] Fix js/http-to-file-access: validate Content-Type and size for LFS PDF download by @github-actions[bot] in #41635
• Update gh aw update --org to support workflow-targeted updates and repo prefiltering by @pelikhan with @Copilot in #41617
• panicinlibrarycode: enforce FuncLit boundaries for init/doc panic exemptions by @pelikhan with @Copilot in #41631
• [instructions] Sync instruction files with release 0.81.4 by @github-actions[bot] in #41644
• Replace AWF runtime Python patch with shared JavaScript helper by @pelikhan with @Copilot in #41623
• Add blog post on custom linters, Sergo, Linter Miner, and LintMonster by @pelikhan with @Copilot in #41663
• reduce instructions-janitor max lines target from 500 to 400 by @pelikhan with @Copilot in #41673
• Retry one completed Copilot BYOK proxy auth failure as a fresh run by @pelikhan with @Copilot in #41629
• Make manifest-version optional in aw.yml by @pelikhan with @Copilot in #41687
• Harden API consumption log collection against continuation timeouts by @pelikhan with @Copilot in #41676
• [spec-enforcer] Enforce specifications for setutil by @github-actions[bot] in #41684
• Refactor duplicated key sorting, engine env assembly, and engine max-* codemods by @pelikhan with @Copilot in #41674
• fix(harness): exit 0 when expected safe-outputs already produced despite numerous permission-denied by @pelikhan with @Copilot in #41675
• improve update/upgrade --org: version tags, current version display, unified repo discovery by @pelikhan with @Copilot in #41627
• Fix go-logger preflight manifest generation failing on jq filter quoting by @pelikhan with @Copilot in #41695
• fix(USE-001): add standardized error codes to pr_review_buffer and set_issue_type handlers by @pelikhan with @Copilot in #41692
• Enforce GH_AW_POLICY_STRICT at runtime for non-strict compiled workflows by @pelikhan with @Copilot in #41682
• parser: add schema validation tests for sandbox.agent.sudo (regression guard for #41679) by @dsyme with @Copilot in #41681
• logs: populate message field with actionable guidance when no runs found by @pelikhan with @Copilot in #41693
• docs: resolve spec audit — pkg/intent spec, actionpins Mappings field, linters 4 new subpackages by @pelikhan with @Copilot in #41723
• fix: post-completion MCP relaunch failure is non-fatal; audit-diff surfaces removed denied domains by @pelikhan with @Copilot in #41713
• Normalize report-format guidance across non-compliant agentic workflows by @pelikhan with @Copilot in #41728
• Harden Super Linter workflow against generated-summary linting and log art...

v0.81.4

🌟 Release Highlights

This release focuses on reliability, performance, and new operational capabilities — fixing silent failures across the compiler and agent harnesses, reducing wasted credits on hot paths, and adding important new tooling.

✨ What's New

• --org flag for upgrade command (#41335) — The upgrade command now supports --org and --repos flags, matching the org-wide mode already available in update. Bulk-preview or open upgrade PRs across an entire organization with a single command.

• Daily YAML Lint Fixer workflow (#41574) — A new agentic workflow automatically fixes yamllint violations in generated *.lock.yml files, keeping CI lint checks clean without manual intervention.

• AWF Firewall startup failure detection (#41472) — Firewall proxy startup failures (e.g., DNS not yet resolving at probe time) are now detected and surfaced in the agent failure issue — making previously silent infra failures immediately actionable.

• Spec-driven engine.env allowlist (#41465) — The engine environment variable allowlist is now derived from GetSupportedEnvVarKeys rather than a fragile runtime heuristic, enabling strict: false to be safely removed from smoke workflows.

• Copilot engine launched from ${GITHUB_WORKSPACE} (#41459) — The Copilot engine now starts from the workspace root, enabling APM skill discovery for context-aware agent behaviour.

🐛 Bug Fixes & Improvements

• Silent YAML parse errors fixed (#41577) — Five yaml.Unmarshal call sites in workflow_builder.go were silently discarding errors, producing empty step lists on malformed YAML with no diagnostic. All errors now propagate correctly.

• Copilot SDK hang bounded by idle watchdog (#41572) — After an agent's final tool result, sendAndWait could hang indefinitely until the step timeout killed the runner. A post-completion idle watchdog now bounds this wait, preventing wasted runner minutes.

• Codex harness no longer drains tokens on rate-limit reconnects (#41385) — When Codex hit a TPM rate limit and exhausted reconnect retries, the harness kept retrying unnecessarily. It now recognises exhausted-reconnect exits and stops, preserving credits.

• Claude harness stops retrying on max_runs_exceeded (#41361) — Fresh-run fallbacks no longer burn the full quota when a session has already hit its 20-invocation limit.

• Issue Monster false-positive 429 detection eliminated (#41471) — Rate-limit false positives no longer trigger spurious failure handling.

• assign_to_agent no longer posts error comments on PRs (#41475) — Error comments are now only posted to issues, matching expected behaviour.

• Daily schedule runs restored (#41362) — Daily schedule runs broken since June 5 are now fixed.

• workflow_call permissions use union of caller + worker (#41387) — Imported workflow_call permissions in generated call jobs are now correctly annotated.

⚡ Performance

• YAML generation is faster (#41333) — Duplicate run-script scans in the skip-validation fast path collapsed to a single pass, restoring compilation performance.

• Design Decision Gate costs reduced (#41332) — Now defaults to Sonnet instead of Opus and skips issue lookups on no-op paths, meaningfully reducing AI credit consumption per run.

🔒 Security

• Safe-output detection hardened (#41547) — Detection stays in warn mode on parser/agent failures, ensuring non-reviewable safe outputs are blocked rather than passed through silently.

• Pi threat-detection model normalisation (#41545) — Pi threat-detection models are now normalised before Copilot fallback, preventing misclassification due to model name variance.

🔧 Internal

• Bumped gh-aw-firewall to v0.27.11 and regenerated pinned workflow artifacts (#41555)
• Bumped Codex 0.142.1 and Copilot SDK 1.0.4 (#41430)
• Extracted shared org-wide runner for update and upgrade commands (#41553)

Generated by 🚀 Release · 44.1 AIC · ⊞ 8.3K

---

What's Changed

• [community] Update community contributions in README by @github-actions[bot] in #41359
• Lower Design Decision Gate runtime cost with Sonnet default and conditional issue lookups by @pelikhan with @Copilot in #41332
• Add GitHub lockdown guard-policy warning and sync security specs by @pelikhan with @Copilot in #41331
• Reduce YAMLGeneration time by collapsing duplicate run-script scans by @pelikhan with @Copilot in #41333
• fix(go-logger): restore daily schedule runs broken since June 5 by @pelikhan with @Copilot in #41362
• feat: add --org to upgrade command by @pelikhan with @Copilot in #41335
• Surface assign-to-agent auth/availability failures in agent failure issues/comments by @pelikhan with @Copilot in #41336
• chore: set sandbox.agent.sudo: false on 30% of agentic workflows by @pelikhan with @Copilot in #41380
• [ubuntu-image] research: update Ubuntu runner image analysis for 20260622.220.1 by @github-actions[bot] in #41373
• [code-scanning-fix] Fix workflow-graphql-static-concat: extract GraphQL query to named constant by @github-actions[bot] in #41357
• fix(manualmutexunlock): distinguish struct instances sharing a mutex field by @pelikhan with @Copilot in #41383
• Add nolint parity to non-enforced context-family linters by @pelikhan with @Copilot in #41382
• Fix stale frontmatter hash in design-decision-gate.lock.yml by @pelikhan with @Copilot in #41390
• Enable Copilot assignment permissions in Issue Monster workflow by @pelikhan with @Copilot in #41389
• [jsweep] Clean upload_assets.cjs by @github-actions[bot] in #41371
• fix: improve agent assignment failure issue reports by @pelikhan with @Copilot in #41392
• [spec-extractor] Update package specifications for tty, types, typeutil, workflow by @github-actions[bot] in #41414
• fix: stop codex harness retry loop draining tokens on exhausted rate-limit reconnects by @pelikhan with @Copilot in #41385
• refactor: consolidate triplicate merge helpers and add sliceutil.SortedKeys by @pelikhan with @Copilot in #41388
• fix: use union of caller + worker permissions for call-workflow jobs by @pelikhan with @Copilot in #41387
• fix: revert sandbox.agent.sudo to true in glossary-maintainer workflow by @pelikhan with @Copilot in #41426
• Handle known Windows ConPTY runner failure in daily CLI integration by @pelikhan with @Copilot in #41428
• fix(firewall): skip non-Squid diagnostic lines in generate_usage_activity_summary by @pelikhan with @Copilot in #41429
• chore: bump Codex 0.142.1, Copilot SDK 1.0.4 by @pelikhan with @Copilot in #41430
• Send GraphQL-Features header for issue intent mutations when issue_intents runtime feature is enabled by @pelikhan with @Copilot in #41425
• Issue Monster: eliminate false-positive 429 rate-limit detection by @pelikhan with @Copilot in #41471
• Standardize CLI help section order and GHE note wording by @pelikhan with @Copilot in #41461
• feat: spec-driven engine.env allowlist via GetSupportedEnvVarKeys; remove strict: false from smoke workflows by @pelikhan with @Copilot in #41465
• Annotate imported workflow_call permissions in generated call jobs by @pelikhan with @Copilot in #41464
• fix: prevent assign_to_agent from posting error comments on PRs by @pelikhan with @Copilot in #41475
• Reduce sandbox.agent.sudo: false usage to policy target across agentic workflows by @pelikhan with @Copilot in #41463
• Update existing MCP configs with required gh-aw server fields by @pelikhan with @Copilot in #41462
• feat: detect AWF firewall startup failures and surface them in the agent failure issue by @pelikhan with @Copilot in #4...

v0.81.3

🌟 Release Highlights

This release focuses on expanded automation reach with org-wide update management, greater expressiveness through GitHub Actions expression support in more places, and a round of critical fixes across Windows, rootless installs, and assignee resolution.

⚠️ Breaking Changes

sandbox.agent.network-isolation renamed to sandbox.agent.default-route

The frontmatter key sandbox.agent.network-isolation has been renamed to sandbox.agent.default-route (#41302). Update any workflows using this key to use the new name.

✨ What's New

• Organization-wide gh aw update — Run gh aw update across an entire org with dry-run PR previews before applying changes, making fleet-wide workflow upgrades safer and more auditable (#41247).
• Templatable safe-outputs.staged values — safe-outputs.staged now accepts GitHub Actions expressions (${{ ... }}), enabling dynamic output values at workflow runtime (#41296).
• link-sub-issue accepts GitHub expressions — The allowed-repos field in link-sub-issue now supports GitHub Actions expressions for more flexible cross-repo linking (#41237).
• ready_for_review trigger support — pull_request_target workflows can now trigger on the ready_for_review event, enabling automation when draft PRs are marked ready (#41161).
• GH_HOST support in gh aw trial — gh aw trial --clone-repo now correctly honors the GH_HOST environment variable for GHES environments (#41159).
• Sudo enabled in agentic sandboxes — All agentic workflow sandboxes now have sudo available by default, unblocking common agent install patterns (#41313).
• Firewall v0.27.10 + mcpg v0.3.30 — Network-isolated workflows omit unnecessary sudo from generated lock files; bundled firewall and MCP gateway updated (#41269).

⚡ Performance

• Parallelized audit analysis — gh aw audit now runs analysis tasks in parallel, significantly reducing latency for long-running workflows (#41185).

🐛 Bug Fixes

• Windows ConPTY crash fixed — Removed a compat import that caused gh aw to crash on startup on Windows (#41235).
• Rootless AWF install — gh aw installs correctly into $HOME/.local without root and properly exports $GITHUB_PATH in rootless environments (#41310).
• Copilot assignee resolution restored — Assignee checks now prefer issue-scoped resolution, fixing cases where the wrong user was assigned (#41306).
• UpdateContainerPins no longer wipes containers — Fixed a regression where gh aw update erased the entire containers section on every run (#41262).
• Locked-PR 422 handled gracefully — Safe outputs now treats HTTP 422 on locked PRs as a soft skip with retry rather than a hard failure (#41155).
• Compiler error quality improved — Errors now include accurate YAML context offsets, import hints, and early engine validation to help authors fix issues faster (#41234).
• set_issue_type migrated to REST API — Replaced the GraphQL-based set_issue_type safe output with a single REST call for better reliability (#41241).
• Linter fixes — lenstringsplit false positives with empty separators and ctxbackground false negatives in closures are resolved (#41188, #41187).
• Codex MCP CLI wrapper resolution — Fixed safe output path resolution for the Codex MCP CLI wrapper (#41242).

📚 Documentation

• Safe rollout guidance streamlined for clarity (#41272).
• Glossary updated with latest terminology (#41211).

Generated by 🚀 Release · 36.2 AIC · ⊞ 8.3K

---

What's Changed

• Remove redundant python-dataviz imports from daily reporting workflows by @pelikhan with @Copilot in #41158
• Support ready_for_review for pull_request_target triggers by @pelikhan with @Copilot in #41161
• fix: treat locked-PR 422 as soft skip with retry in safe_outputs by @pelikhan with @Copilot in #41155
• Add SEC-005 exemption for issue_intents.cjs false positive by @pelikhan with @Copilot in #41182
• Allow AgentRx native package installs in Daily AgentRx Trace Optimizer by @pelikhan with @Copilot in #41183
• Pin Daily Sub-Agent Model Resolution Audit sub-agent to a valid Codex model by @pelikhan with @Copilot in #41184
• Pin RGS-007 workflow action refs to immutable SHAs by @pelikhan with @Copilot in #41189
• [docs] Update glossary - daily scan by @github-actions[bot] in #41211
• Add replace-label to Safe Output Mapping in agentic workflow designer SKILL.md by @pelikhan with @Copilot in #41190
• Honor GH_HOST in gh aw trial --clone-repo repository URLs by @pelikhan with @Copilot in #41159
• fix(lenstringsplit): empty raw-string separator escapes non-empty guard — false positive + wrong autofix by @pelikhan with @Copilot in #41188
• ctxbackground: fix false negative and unsafe autofix for closures by @pelikhan with @Copilot in #41187
• Add explicit permissions to error-message-lint workflow by @pelikhan with @Copilot in #41233
• [jsweep] Clean update_entity_helpers.cjs by @github-actions[bot] in #41166
• fix(footer): render correct trigger type in attribution suffix by @pelikhan with @Copilot in #41186
• Fix false negatives in docs npm update detection by @pelikhan with @Copilot in #41240
• fix: normalize report formatting for daily-rendering-scripts-verifier.md by @pelikhan with @Copilot in #41245
• perf: parallelize audit analysis tasks to cut latency for long-running workflows by @pelikhan with @Copilot in #41185
• refactor(workflow): split threat_detection.go (1542 lines) into focused modules by @pelikhan with @Copilot in #41231
• feat: link-sub-issue allowed-repos accepts GitHub Actions expressions by @pelikhan with @Copilot in #41237
• Migrate set_issue_type safe output from GraphQL to single REST issues.update call by @pelikhan with @Copilot in #41241
• ci: align build-wasm node setup to setup-node@v6 / Node 24 by @pelikhan with @Copilot in #41243
• feat: omit sudo from generated lock.yml when network-isolation is enabled; bump firewall to v0.27.10 and mcpg to v0.3.30 by @lpcox with @Copilot in #41269
• [review] Migrate set_issue_type safe output from GraphQL to single REST issues.update call by @github-actions[bot] in #41284
• [docs] docs: unbloat safe rollout guidance by @github-actions[bot] in #41272
• fix(windows): remove compat import to prevent ConPTY startup crash by @pelikhan with @Copilot in #41235
• Fix compiler error quality: YAML context offset, import hint, early engine validation by @pelikhan with @Copilot in #41234
• [docs] Update Astro dependencies - 2026-06-24 by @github-actions[bot] in #41258
• [linter-miner] feat(linters): add stringreplaceminusone linter by @github-actions[bot] in #41285
• fix: UpdateContainerPins wipes containers section on every gh aw update run by @pelikhan with @Copilot in #41262
• Fix Codex MCP CLI wrapper resolution for safe outputs by @pelikhan with @Copilot in #41242
• [rendering-scripts] Render Codex experimental JSONL logs in the run step summary by @github-actions[bot] in #41201
• build(deps-dev): Bump starlight-github-alerts from 0.2.0 to 0.3.0 in /docs by @dependabot[bot] in #41000
• build(deps): Bump sharp from 0.35.1 to 0.35.2 in /docs by @dependabot[bot] in #40999
• [caveman] Optimize instruction verbosity — agentic-chat, campaign, developer.instructions (2026-06-24) by @github-actions[bot] in #41301
• Reduce ambient-context bloat in shared Python dataviz import by @pelikhan with @Copilot in #41304
• fix: use python3 -m pip install for ...

v0.81.2

🌟 Release Highlights

This release introduces a new observability workflow for model resolution auditing, ships updated firewall and MCP gateway defaults, and improves internal GitHub MCP rendering consistency.

✨ What's New

• Daily Model Resolution Audit — A new daily-model-resolution workflow detects model size mismatches in inlined sub-agents by cross-referencing workflow declarations against api-proxy event logs. It runs on a daily schedule (or on demand) and opens a scoped daily issue with a per-workflow, per-sub-agent table showing declared alias vs. observed model vs. resolution status (✅ / ⚠️ / ❓). This makes it much easier to catch accidental large-model usage where a small model was intended. (#41130)

🐛 Bug Fixes & Improvements

• Integration test stability — Fixed a failing Integration: Workflow Features GitHub Actions job to restore CI reliability. (#41153)
• GitHub MCP serialization refactor — Consolidated duplicated map-serialization and section-closing logic across local, remote, and custom GitHub MCP rendering paths into shared helpers. Token passthrough, GITHUB_HOST/toolset propagation, and guard-policy footer handling are now consistently applied across all modes, reducing the risk of future drift. (#41038)

🔧 Dependency Updates

• Default firewall & MCP gateway bumped — gh-aw-firewall default updated to v0.27.9 and gh-aw-mcpg to v0.3.29. Compiled workflows now reference new immutable container digests for the AWF agent, squid, api-proxy, cli-proxy, and MCP gateway. (#41129)
• @types/node 25 → 26 — TypeScript type declarations in actions/setup/js updated to the Node.js v26 API surface. Dev-only; no runtime impact. (#40998)

Generated by 🚀 Release · 28.1 AIC · ⊞ 8.3K

---

What's Changed

• Refactor GitHub MCP serialization into shared section helpers by @pelikhan with @Copilot in #41038
• feat: daily sub-agent model resolution audit workflow by @pelikhan with @Copilot in #41130
• build(deps-dev): Bump @types/node from 25.9.2 to 26.0.0 in /actions/setup/js by @dependabot[bot] in #40998
• Bump default gh-aw-firewall to v0.27.9 and gh-aw-mcpg to v0.3.29 by @lpcox with @Copilot in #41129
• [WIP] Fix failing GitHub Actions job 'Integration: Workflow Features' by @pelikhan with @Copilot in #41153

Full Changelog: v0.81.1...v0.81.2

v0.81.1

🌟 Release Highlights

This release focuses on reliability and resilience — fixing common workflow failure modes, hardening remote fetch operations, and expanding platform support with a new network-isolation topology for ARC/DinD runners.

✨ What's New

• Network-isolation topology mode for ARC/DinD runners (#41088): Compiler now emits the --network-isolation topology for AWF workflows, enabling egress control without requiring NET_ADMIN or host iptables. This unblocks deployments on Actions Runner Controller (ARC) with Docker-in-Docker.

• issue_intents runtime support (#41092): Issue mutations can now carry structured rationale, confidence, and suggest metadata when the server supports the issue_intents feature flag, with automatic fallback to the legacy format — no workflow changes required.

• Automatic threat-detection engine fallback for pi (#41098): When a workflow targets the pi engine (which does not support threat-detect directly), the compiler now automatically defaults to Copilot for threat detection — eliminating the need to configure this per workflow.

🐛 Bug Fixes & Improvements

• Safe-outputs graceful no-op (#41037): Fixes a leading cause of workflow failures where a missing or empty safe-outputs file caused safe_outputs job failures. The runtime now treats a missing file as an empty collection and continues cleanly.

• Improved assign-to-agent diagnostics (#41047): Copilot assignee lookup during safe-outputs now searches across all known Copilot bot aliases, preventing silent resolution failures that blocked agent assignment.

• Hardened go-gh remote fetch (#41036): Contents API path construction is now properly escaped, REST clients are bounded, and client creation logic is deduplicated — reducing reliability gaps on remote repository operations.

• Accurate audit pre-activation diagnostics (#41022): gh aw audit no longer incorrectly reports "failed before agent activation" for runs that did execute but exported no telemetry. Diagnosis now correctly distinguishes pre-activation failures from agent-run failures. See audit reference.

📚 Documentation

• Editing workflows guide streamlined (#41086): Dense sub-sections replaced with readable inline prose for faster scanning.
• Release cadence clarified (#41087): Docs now reflect the weekly/bi-weekly minor release clock (not best-effort semver).

Generated by 🚀 Release · 48.4 AIC · ⊞ 8.3K

---

What's Changed

• audit: distinguish pre-activation failures from agent-run failures without telemetry by @pelikhan with @Copilot in #41022
• [docs] docs: unbloat editing workflows guide by @github-actions[bot] in #41086
• Improve assign-to-agent diagnostics for Copilot assignee alias lookup by @pelikhan with @Copilot in #41047
• docs: release cadence is weekly/bi-weekly minor, not best-effort semver by @pelikhan with @Copilot in #41087
• [linter-miner] linter: add lenstringsplit analyzer by @github-actions[bot] in #41090
• Harden go-gh remote fetch callsites with escaped contents paths/refs and bounded REST clients by @pelikhan with @Copilot in #41036
• Default threat detection to Copilot when engine is pi by @pelikhan with @Copilot in #41098
• Add compiler support for AWF --network-isolation topology mode (ARC/DinD-compatible egress) by @lpcox with @Copilot in #41088
• [caveman] Optimize instruction verbosity — adr-writer.agent.md, workflow-constraints.md (2026-06-23) by @github-actions[bot] in #41105
• Add issue_intents runtime support for issue labels, type, and fields by @pelikhan with @Copilot in #41092
• fix: treat missing safe-outputs file as empty collection (graceful no-op) by @pelikhan with @Copilot in #41037

Full Changelog: v0.81.0...v0.81.1

v0.81.0

🌟 Release Highlights

This release delivers expanded engine configurability, a batch of high-impact bug fixes across the compiler and CLI, and tighter security defaults — making workflows more reliable and cost-efficient out of the box.

⚠️ Breaking Changes

• CLI flag rename (#40822): The --skip-secret, --disable-security-scanner, and --disable-release-bump flags have been renamed to the --no-secret, --no-security-scanner, and --no-release-bump convention. Update any scripts or CI configurations that pass these flags.

✨ What's New

• Model provider selection (#40903): Workflows can now declare an engine-level model-provider field, enabling provider-aware runtime wiring — including running Claude models via the Copilot engine. A new Claude-on-Copilot smoke workflow validates the path end-to-end.
• engine.driver field (#40897): A new shared engine.driver field allows specifying the Pi agent core as a built-in driver, simplifying engine configuration.
• sandbox.agent.platform frontmatter (#40877): Workflows can now read the AWF platform type via sandbox.agent.platform, making platform-conditional logic easier to express.
• GH_AW_RUNTIME_FEATURES flag support (#40824): Built-in workflow jobs now honor the GH_AW_RUNTIME_FEATURES environment variable for progressive feature rollout.
• replace-label safe-output type (#40423): A new experimental replace-label safe-output type lets workflows atomically swap issue/PR labels.
• Smarter merge-pull-request (#40915): The merge-pull-request safe-output now refuses to merge when the target branch has no upstream open PR, preventing accidental merges into stale branches.
• Slash-command status tracking (#40928): Centralized slash-command status comments are now updated with a link to the dispatched workflow run, so you always know where the work went.
• Firewall activity backfill (#41046): Activity reports can now be backfilled from usage artifact domain aggregates, filling gaps in firewall observability data.

🐛 Bug Fixes & Improvements

• Compiler: fix hyphenated workflow_call inputs (#41019): Workflows with hyphenated input names in workflow_call events now compile correctly.
• Compiler: fix on.bots/on.roles state leakage (#41018): A state leak that could corrupt workflow_run trigger generation when on.bots or on.roles were set has been fixed.
• Compiler: reduce template-injection noise (#41025): The compiler now emits fewer false-positive template-injection diagnostics in generated workflow run blocks.
• Linter: wgdonenotdeferred loop-scope fix (#41026): The linter no longer incorrectly flags deferred WaitGroup.Done() calls inside goroutine closures in loop bodies.
• Status comments: correct run URL and name (#40831): Centralized repo status comments now display the actual workflow name and a working link to the run.
• Detection: honor continue-on-error (#40790): Detection jobs now respect continue-on-error when the external detector result file is missing.
• Detection: always run on safe outputs (#40882): Cache-memory detection is now guaranteed to run even on noop and boop safe-output types.
• Ollama reachability (#40888): Ollama is now bound to 0.0.0.0 so the api-proxy Docker container can reach it correctly.
• Action ref pass-through (#40887): Supported local and Docker action refs are now passed through without triggering pin resolution.
• --engine log filter (#40793): gh aw logs --engine now correctly filters runs by engine ID.
• Telemetry: Copilot/Pi turn and token reporting (#40792): Turn count and token usage are now emitted correctly for the Copilot and Pi engines.
• Portfolio analyst turn runaway (#40858): The portfolio analyst workflow now has max-turns and max-ai-credits guardrails to prevent unbounded execution.

⚡ Performance

• ~9,600 chars trimmed from ambient context (#40874): Per-run token usage is meaningfully reduced by pruning verbose ambient context that agents did not need.
• Failure investigator cost reduction (#40880): The aw-failure-investigator workflow has been optimized to use fewer AI credits per run.

🔒 Security

• Hardened checkout credential defaults (#40794): Generated checkout steps now default to persist-credentials: false, reducing the credential exposure window in generated workflows.

📚 Documentation

• max-daily-ai-credits cost warnings (#41055): Docs now warn about the API unit cost implications of max-daily-ai-credits and include examples for disabling the field, with a recommendation to configure at the org level.
• Authoring guidance improvements (#40819, #40820): Tightened guidance on noop semantics, workflow_run triage, report window scoping, and permissions documentation.

Generated by 🚀 Release · 36.8 AIC · ⊞ 8.2K

---

What's Changed

• [docs] Update glossary - weekly full scan by @github-actions[bot] in #40776
• Sync workflow designer skill with safe-output and network reference docs by @pelikhan with @Copilot in #40742
• Reduce tool-denial failures in Daily Compiler Threat Spec Optimizer by @pelikhan with @Copilot in #40743
• Align threshold linters with sibling analyzer behavior by @pelikhan with @Copilot in #40740
• fix(seenmapbool): eliminate duplicate diagnostics for set-maps in closures by @pelikhan with @Copilot in #40741
• [architecture] Update architecture diagram - 2026-06-22 by @github-actions[bot] in #40769
• [schema-coverage] feat: Add schema coverage demo for max-turn-cache-misses field by @github-actions[bot] in #40753
• safe-outputs: make create-pull-request bundling resilient to diverged-history replay conflicts by @pelikhan with @Copilot in #40720
• [code-scanning-fix] Fix workflow-graphql-id-unescaped: eliminate string interpolation in GraphQL mutations via variables by @github-actions[bot] in #40757
• [code-simplifier] simplify: dedup validateAllowedIssueFields and extract parseUnknownModelAICreditsFromAuditEntry by @github-actions[bot] in #40725
• fix: increase timeout-minutes from 10 to 30 in example-workflow-analyzer by @pelikhan with @Copilot in #40788
• [jsweep] Clean validate_lockdown_requirements_templates.cjs by @github-actions[bot] in #40730
• fix: resolve SEC-004 false-positive sanitization flags on assign_agent_helpers and update_pull_request by @pelikhan with @Copilot in #40791
• fix: update actions toolset to consolidated tool names (actions_get/actions_list) by @pelikhan with @Copilot in #40787
• refactor: centralise symlink guard into shared lstatGuard helper by @pelikhan with @Copilot in #40795
• fix: restore AIC data in usage-only log collection by @pelikhan with @Copilot in #40786
• pi: propagate GH_AW_NODE_BIN into AWF execution path for Daily News by @pelikhan with @Copilot in #40789
• docs(permissions): move read-only constraint note and add group descriptions by @pelikhan with @Copilot in #40820
• Docs: tighten workflow authoring guidance for noop semantics, workflow_run triage, and report window scoping by @pelikhan with @Copilot in #40819
• fix(logs): --engine filter correctly filters runs by engine ID by @pelikhan with @Copilot in #40793
• fix(telemetry): emit turns/tokens for Copilot/Pi; trim cli-version-checker prompt; foreground recompile by @pelikhan with @Copilot in #40792
• Add GH_AW_RUNTIME_FEATURES support to built-in workflow jobs by @pelikhan with @Copilot in #40824
• fix(cli): rename --skip-secret, --disable-security-scanner, --disable-release-bump to --no-* convention; fix docs gaps by @pelikhan with @Copilot in #40822
• refactor(cli): extract ReportProvenance f...