v1.0.0

🚀 Features

• (app) Check runtime --env checks the env's effective platforms - (74ae0e1) by @bchatard

  RuntimeCheck's Platforms section only ever read cfg.Release.Platforms,
  so heraut check runtime --env <env> (and bare heraut check --env <env>) silently ignored an environment's release.platforms override.

  RuntimeCheck gains an env parameter and resolves the effective platform
  list with the same replace semantics as buildReleasePipelineConfig and
  hasEffectivePlatforms (T89): the env's release.platforms replaces the
  root list wholesale when non-empty, otherwise the root list is checked.
  internal/cmd/check.go reads the root --env flag and threads it through
  runRuntimeCheck for both the bare check and check runtime commands.

  Roadmap:docs/tasks/roadmap.md -> T100 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (config) Validate tag_pattern requires git-cliff generator - (bcd9467) by @bchatard

  tag_pattern is a git-cliff-only field — communique and cocogitto
  silently ignore it. Gate it in validateContentDriver the same way
  tickets is gated, so a non-git-cliff generator paired with tag_pattern
  now fails heraut check config with an actionable hint instead of being
  silently dropped.

  The single check covers top-level changelog/release.notes and their
  per-env equivalents, since MergeContentDriver (ADR-0019) already
  resolves each per-env driver to its effective (generator, tag_pattern)
  pair before validation runs.

  Roadmap:docs/tasks/roadmap.md -> T97 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (scaffold) Warn before dropping unsupported settings on init update - (1a0f3f9) by @bchatard

  The "Update it?" flow in heraut init regenerates .heraut.yml from
  ConfigToAnswers/GenerateYAML, which don't round-trip tickets,
  remote_metadata, release.assets, per-platform base_url/draft/prerelease,
  or per-env changelog/release overrides. Previously this loss was silent.

  Add scaffold.DroppedFields to detect any of these in the loaded config
  and print a warning (with the affected YAML paths) before the wizard
  runs, so the user can decide whether to continue. Export
  config.DefaultBaseURL so the base_url check can ignore type defaults
  filled in by normalizePlatforms.

  Roadmap:docs/tasks/roadmap.md -> T99 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

🐛 Bug Fixes

• (cmd) Unify --version override validation across release/changelog - (e269720) by @bchatard

  heraut release rejected valid CalVer and pre-release --version overrides
  via a SemVer-only regex, while heraut changelog --version validated
  nothing. NewResolver already treats --version as an opaque string with
  zero shape validation (StaticResolver), and spec 03 documents no shape
  restriction.

  Drop the SemVer regex and add tagfmt.ValidateVersionOverride /
  app.ValidateVersionOverride — non-empty, no whitespace, otherwise
  format-agnostic (mirrors ValidateBuildID minus the "/" ban, since a full
  tag override may contain one). Apply identically to both commands.

  Roadmap:docs/tasks/roadmap.md -> T90 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (cmd) Align check/version validation, no-config, and exit codes - (d2406b6) by @bchatard

  Three consistency fixes:

  • version next/current now run config.Validate before resolving, so
    per-env misconfigs (e.g. promotion cycles) print the same path/hint
    output as check config and exit Config, instead of surfacing as raw
    resolver errors deep in the perenv promotion logic.

  • Bare heraut check degrades like check runtime when no config
    file exists (warns and proceeds with cfg = nil) instead of hard
    failing, so it works the same as running each subcommand
    individually in a config-less directory.

  • Bare heraut check now classifies its exit code by what failed:
    Config if the config section had errors (even if runtime/cliff also
    failed), Runtime otherwise.

  Roadmap:docs/tasks/roadmap.md -> T98 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (cmd/release) Inherit root platforms when env release omits platforms - (1b5283e) by @bchatard

  hasEffectivePlatforms returned false whenever an env's release: block
  was non-nil but set no platforms (e.g. notes-only override), even
  though buildReleasePipelineConfig correctly falls back to the root
  platform list in that case. Align the guard with the builder's merge
  semantics: env platforms replace root only when non-empty.

  Roadmap:docs/tasks/roadmap.md -> T89 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (pipeline) Push only the resolved tag, not all local tags - (8e08976) by @bchatard

  Both pipelines ran git push origin --tags, which publishes every local
  tag including stale or experimental ones. Add gitHelper.pushTag(tag) and
  use it to push result.Tag explicitly. Align the changelog pipeline's
  "Push tags" step naming with the release pipeline's "Push tag", and have
  dry-run output name the exact tag being pushed.

  Roadmap:docs/tasks/roadmap.md -> T93 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (platforms/gitlab) Inject token_env into release-time glab calls - (bd0bcb9) by @bchatard

  CreateRelease and UploadAssets passed only hostEnv(), so a custom
  token_env (or a self-hosted instance with its own token, ADR-0025's
  motivating scenario) would pass heraut check runtime but fail at
  glab release create/upload time. Mirror checkAPIAuth's and
  GitHub's tokenEnvSlice()+hostEnv() composition.

  Roadmap:docs/tasks/roadmap.md -> T88 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (versioning/perenv) Skip pre-release tags in semver-per-env BumpAuto - (938b911) by @bchatard

  tagfmt.ParseVersion's {version} capture is a greedy .+ with no
  pre-release filter, so a tag like dev/1.3.0-rc.1 parses to bare
  "1.3.0-rc.1" unfiltered. Without versionsort.suffix, git can sort that
  tag above dev/1.2.3, making it currentVersion — and BumpVersion's
  strconv.Atoi on the "0-rc.1" patch segment fails, the same "invalid
  patch" mode T92 fixed for plain semver.

  Export semver.IsBareVersion and apply it in perenv.resolveAuto's
  shared tag-filter loop. The loop is shared with calver-per-env, but
  IsBareVersion accepts any 3-part all-integer version and
  calver.BumpFromDate already does its own per-tag skip, so this is a
  no-op there.

  Roadmap:docs/tasks/roadmap.md -> T103 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (versioning/semver) Anchor breaking-change bang to commit type prefix - (f4a8e9e) by @bchatard

  isBreaking matched "!:" anywhere in the subject, so a fix commit whose
  description happens to contain "!: " (e.g. "fix: handle the foo!: token")
  triggered a spurious major bump. Anchor the bang to the conventional-commit
  type/optional-scope prefix instead.

  Also recognize the hyphenated BREAKING-CHANGE: footer as a synonym of
  BREAKING CHANGE:, per Conventional Commits 1.0.

  Roadmap:docs/tasks/roadmap.md -> T91 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

• (versioning/semver) Skip pre-release tags when resolving current version - (3e4b71a) by @bchatard

  Without versionsort.suffix configured, git's default version:refname tag
  sort orders a pre-release tag (e.g. v1.3.0-rc.1) above its release
  (v1.2.3), so resolveAuto picked the pre-release as the current tag and
  BumpVersion failed parsing its "-rc.1" suffix.

  resolveAuto now walks the sorted tags and picks the first whose bare form
  is a plain MAJOR.MINOR.PATCH (new isBareVersion helper), mirroring the
  CalVer resolver's skip-unparsable-tag pattern. Falls back to
  initial_version if every matching tag is non-conforming. Documented in a
  new "Pre-release tags" subsection of spec 04.

  Also adds T102 (spec 04's bump-determination table doesn't reflect T91's
  anchored "!" / BREAKING-CHANGE: synonym) and T103 (semver-per-env's
  BumpAuto may hit the same pre-release-tag issue via perenv.resolveAuto)
  as follow-up roadmap tasks discovered during this work.

  Roadmap:docs/tasks/roadmap.md -> T92 Co-Authored-By:Claude Fable 5 noreply@anthropic.com

🚜 Refactor

• (config) Extract shared EffectivePlatforms helper (T105) - (f9ec386) by @bchatard

  internal/app/pipeline.go (buildReleasePipelineConfig), internal/cmd/release.go
  (hasEffectivePlatforms, T89), and internal/app/check.go (RuntimeCheck, T100)
  each independently implemented the same env/root release-platforms merge
  rule. config.EffectivePlatforms is now the single implementation; the cmd
  and app call sites delegate to it.

  Roadmap:docs/tasks/roadmap.md -> T105

• Hygiene pass — drop T-id comments, dedupe iota, unify case - ([8413f47](https://github.com/adaouat/he...

v0.31.0

🚀 Features

• (app) One heraut check runtime Platforms row per configured entry - (6bdc4b4) by @bchatard

  RuntimeCheck now dispatches one Platforms row per release.platforms
  entry, labeled by the entry's configured name, running each entry's
  full Check() (binary + token + project/repository + API auth). With
  no platforms configured, falls back to a binary-only probe of glab
  and gh. Removes configuredPlatforms/findPlatformCfg, which assumed at
  most one entry per platform type.

  Roadmap:docs/tasks/roadmap.md → T86

• (config) Require unique platform name, allow self-hosted base_url - (2d6f2f6) by @bchatard

  Adds a required name field to release.platforms entries (unique per
  list scope) and lifts ADR-0020's "self-hosted hosts are not yet
  supported" gate on base_url — both are prerequisites for publishing to
  two instances of the same platform type (ADR-0025).

  Roadmap:docs/tasks/roadmap.md → T83

• (platforms/github) Support self-hosted instances via GH_HOST - (c113508) by @bchatard

  Name() now returns the configured platform name, ReleaseURL() honors
  base_url, and CreateRelease/UploadAssets/Check inject GH_HOST and
  GH_ENTERPRISE_TOKEN when base_url targets a non-default (GHES) host.
  Actions-token autologin is skipped for self-hosted instances in favor
  of the configured token.

  Roadmap:docs/tasks/roadmap.md → T85

• (platforms/gitlab) Support self-hosted instances via GITLAB_HOST - (3bfdb06) by @bchatard

  Name() now returns the configured platform name, ReleaseURL() honors
  base_url, and CreateRelease/UploadAssets/Check inject GITLAB_HOST when
  base_url targets a non-default host. CI autologin is skipped for
  self-hosted instances in favor of the configured token.

  Roadmap:docs/tasks/roadmap.md → T84

📚 Documentation

• (adr) Add ADR-0025, supersede ADR-0020 (multi-instance platforms) - (b3ae621) by @bchatard

  ADR-0025 documents the lifted base_url gate, per-platform CLI host
  targeting (GH_HOST/GITLAB_HOST/GH_ENTERPRISE_TOKEN), the required
  unique name field, and the restructured heraut check runtime
  Platforms section delivered by T83-T86. Spec 05 gains the self-hosted
  / multi-instance subsection and the name field in both platform
  examples.

  Roadmap:docs/tasks/roadmap.md → T87

• (app) Clarify Platforms section dispatch order in RuntimeCheck - (e47b250) by @bchatard

  The previous comment described the old fixed glab→gh dispatch order,
  which T86 replaced with one row per configured release.platforms entry
  (falling back to the binary-only probe only when none are configured).

• (plans) Add multi-instance platforms implementation plan - (4f9eb70) by @bchatard

  Implementation plan for T83-T87 (ADR-0025, multi-instance
  same-platform releases), executed via subagent-driven development.

  Roadmap:docs/tasks/roadmap.md → T83-T87

• Design spec for multi-instance same-platform releases - (2b3e99b) by @bchatard

  Adds a future-ideas entry and design spec covering name-based platform
  disambiguation, lifting the ADR-0020 base_url gate, and per-invocation
  GH_HOST/GITLAB_HOST targeting for self-hosted instances.

🎨 Styling

• (app) Drop unneeded loop variable copy in check.go - (60b3e9d) by @bchatard

  Go 1.22+ gives each for-range iteration its own variable, so the
  bin := bin copy added by T86 was redundant.

Commit Statistics

• 9 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 9 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.

v0.30.2

🐛 Bug Fixes

• (deps) Bump forge to v0.16.0 - (56d38f7) by @bchatard

  Pulls in forge's whatsnew rendering fixes: dropped glamour's
  terminal-querying "auto" style (which fell back to raw markdown when
  the terminal didn't answer), added $PAGER support, and fixed a
  double-print bug when the pager exits non-zero.

Commit Statistics

• 1 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 1 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.
• 1 day(s) passed between releases.

v0.30.1

🎨 Styling

• Gofmt -s on calver parser_test - (e5ec535) by @bchatard

  Struct-field alignment that gofmt -s wants but golangci-lint (CI) doesn't enforce —
  the lone pre-existing go_fmt failure surfaced by 'hk check --all', a prerequisite for
  adding the hk lint gate to CI.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

⚙️ Miscellaneous Tasks

• Run the full hk linter set in CI (hadolint, actionlint, typos, …) - (0906de5) by @bchatard

  The reusable ci job only runs golangci-lint; hk's other linters (hadolint,
  actionlint, yamlfmt, typos, pkl, go fmt) were enforced only by the local pre-commit
  hook (staged files, bypassable). Add an hk job running 'hk check --all --check
  --skip-step golangci_lint' so they're enforced server-side. hk's per-step file
  selection scopes each linter — hadolint only touches the Dockerfile, so it runs only
  when relevant without a fragile required-check path filter.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• Fix shellcheck findings in release.yml run blocks - (c8cb125) by @bchatard

  The new hk CI job (actionlint -> shellcheck) flagged unquoted variable
  expansions in release.yml. Quote $FRESH_BIN, $GITHUB_ENV, and
  $GITHUB_OUTPUT (SC2086). The imagetools create command intentionally
  relies on word splitting / globbing for multiple -t tags and digest
  files, so suppress SC2046 there with an explanatory comment.

  Co-Authored-By:Claude Sonnet 4.6 noreply@anthropic.com

Commit Statistics

• 3 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 3 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.

v0.30.0

🚀 Features

• (app) Propagate tickets config onto content drivers - (10410fa) by @bchatard

  withEnvDerivations now carries cfg.Tickets onto the changelog/release-notes driver
  clones (the same clone point as remote_metadata), so the gitcliff generator can
  inject link_parsers. Part of T80.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (config) Add tickets field for issue-tracker links - (770c318) by @bchatard

  Top-level tickets list ([]Ticket{Pattern, URL}) parsed from .heraut.yml, plus the
  programmatic ContentDriver.Tickets carrier the app layer will populate. Part of T79.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (config) Validate tickets (regex, {ticket} url, git-cliff only) - (a159801) by @bchatard

  validateTickets:each pattern compiles as a regex, each url is an absolute http(s)
  URL containing {ticket}, and tickets require the git-cliff generator (the only one
  with a link mechanism). Part of T79. Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (config) Add tickets to schema + sample + fixtures - (0de6706) by @bchatard

  Expose the tickets array in schema.json (pattern + url, both required), document
  it in docs/heraut.sample.yml, and add valid + invalid schema fixtures (the invalid
  one omits the required url). Part of T82.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (generators/gitcliff) Inject ticket link_parsers into effective config - (cc1925d) by @bchatard

  effectiveConfig now chains injectLinkParsers after the heading postprocessor: each
  ticket becomes a git-cliff { pattern, href } entry — pattern wrapped in a capture
  group only when it has none (so {ticket}->$1 is the URL value), appended to any
  existing [git].link_parsers. Part of T80.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (generators/gitcliff) Render ticket links in changelog and release notes - (98f83e1) by @bchatard

  Both print_commit macros append commit.links after the PR segment as
  ( TICKET ). Verified against the real git-cliff: subject and footer (Refs:)
  tickets both render with clean whitespace, location-independent. Part of T81.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

📚 Documentation

• (plans) Add ticket-linking design spec - (6dee113) by @bchatard

  Approved brainstorming design: first-class .heraut.yml ticket linking via a
  top-level tickets list, injected into git-cliff's link_parsers (location-
  independent — subject/body/footer), rendered in both changelog and release notes.
  git-cliff only; append-only; {ticket} maps to the first capture group (or full
  match) so label vs URL value can differ (e.g. GH-123 → /issues/123).

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (plans) Add ticket-linking implementation plan - (d33f307) by @bchatard

  Seven TDD tasks (config field, validation, link_parsers injection, driver
  propagation, both templates, schema/sample/fixtures, ADR/spec/roadmap) for T79,
  each with failing test, implementation, and commit steps.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (roadmap) Add Phase 15 — ticket linking (T79–T82) - (dfc44c5) by @bchatard

  New phase tracking the ticket-linking feature: config surface + validation (T79),
  link_parsers injection + driver propagation (T80), template rendering (T81), and
  schema/sample/ADR/spec (T82).

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• ADR-0024 ticket linking; document tickets; complete T79–T82 - (485c07a) by @bchatard

  Record the link_parsers-over-commit_preprocessors decision (ADR-0024), document the
  tickets config in spec 02, and flip Phase 15 (T79–T82) to done with completion notes.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

Commit Statistics

• 10 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 10 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.

v0.29.0

🚀 Features

• (cmd) Add --build to heraut release for build-id release flows - (280b29d) by @bchatard

  Mirror heraut changelog --build: --build requires --version, the value is
  validated via app.ValidateBuildID, and it is passed to NewResolver, which renders
  {build} into the tag and returns a StaticResolver. The release pipeline consumes
  result.Tag unchanged, so no pipeline/resolver change is needed — the build path
  already existed (deferred from T52). Build-id teams can now run a full release
  (changelog -> tag -> publish) per CI build. release-per-build is allowed freely:
  passing both --version and --build is already a deliberate, scripted opt-in.

  Closes T57.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

🐛 Bug Fixes

• (generators/gitcliff) Render New Contributors from any platform's remote - (316ce84) by @bchatard

  The embedded release-notes template only read gitlab.contributors, so non-GitLab
  repos never rendered the New Contributors section. Extract a print_contributors
  macro and call it once over the first-time contributors concat'd across all five
  remote namespaces (gitlab/github/gitea/bitbucket/azure_devops). The single call
  over a merged list keeps git-cliff's proven single-block whitespace — verified
  against the real CLI for the empty, single-platform, and multi-platform cases (no
  blank-line regression).

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

📚 Documentation

• (roadmap) Mark T68 done (checkbox lag; work landed in ADR-0021) - (7061e89) by @bchatard

  T68's decision and PoC were completed and documented in its Done note (the
  context-injection shape resolved in ADR-0021, dependents T69/T71a/T71b already
  done) — only the heading checkbox was never flipped. Flip it. T57 is now the
  single open roadmap task.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (roadmap) Complete T57 (heraut release --build) - (d86c09b) by @bchatard

  Flip T57 to done with a completion note: mirrored changelog --build, no
  pipeline change (resolver already handled build), release-per-build allowed
  freely. Roadmap is now fully resolved — no open tasks.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (specs) Document heraut release --build (T57) - (3e8e37a) by @bchatard

  Flip the {build} support table — heraut release --build is now supported, not
  changelog-only — add the release example, the --build flag to the release command
  reference, and note that release-per-build (one platform release per build) is
  intentional and unguarded.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

🎨 Styling

• (generators/gitcliff) Drop redundant version heading from release notes - (c977f61) by @bchatard

  Release notes are attached to a GitHub/GitLab release whose header already shows
  the tag/version and date, so the '## [version] - date' line in the body was pure
  duplication. Drop it (the changelog template keeps its heading — that file
  accumulates every release). Also pack the macro definitions (whitespace-only).
  Verified against the real git-cliff: config accepted, contributors render across
  all platforms, no output regression (one leading blank line is git-cliff
  structural with an empty header, and invisible on the rendered release page).

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

⚙️ Miscellaneous Tasks

• Keep CHANGELOG.md in Docker build context for the whatsnew embed - (acac7b6) by @bchatard

  changelog.go embeds CHANGELOG.md into the binary (//go:embed, 09fe57c) for the
  whatsnew offline fallback, so the Dockerfile's COPY . . + go build needs the file
  present. The .dockerignore '*.md' rule was excluding it, breaking the image build
  with 'pattern CHANGELOG.md: no matching files found'. Add a !CHANGELOG.md
  exception; other .md files stay ignored.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

Commit Statistics

• 7 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 7 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.
• 1 day(s) passed between releases.

v0.28.0

[0.28.0] - 2026-06-10

🚀 Features

• (whatsnew) Embed CHANGELOG.md as the offline fallback - (09fe57c) by @bchatard

  Adds a root package embedding CHANGELOG.md into a string (the embed must live
  at the repo root to reach the file) and passes it to WhatsNewConfig.Changelog,
  so whatsnew renders the bundled changelog when GitHub is unreachable and the
  cache is cold (forge ADR-0012 tier D).

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

💼 Other

• (deps) Bump forge to v0.15.0 - (93cee93) by @bchatard

  Brings WhatsNewConfig.Changelog (forge ADR-0012 tier D) for the embedded
  changelog offline fallback.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

Commit Statistics

• 2 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 2 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.

v0.27.0

[0.27.0] - 2026-06-10

🚀 Features

• (cmd) Add whatsnew command, adopt forge's shared hint wiring - (9457530) by @bchatard

  Registers updatecheck.WhatsNewCommand (heraut whatsnew shows the release-note
  span newer than the running build), and collapses the update-hint PostRun to a
  single updatecheck.Hinter{…}.PostRun() using the shared CacheFile("heraut") —
  dropping the hand-rolled timeout/cache-path block (forge M11/ADR-0012).

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

💼 Other

• (deps) Bump forge to v0.13.0 - (6a4f278) by @bchatard

  Brings updatecheck.WhatsNewCommand (forge ADR-0012 tier C) for the whatsnew
  wiring, plus the tier A changelog pointer in the upgrade hint.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (deps) Bump forge to v0.14.0 - (b36f93a) by @bchatard

  Brings updatecheck.CacheFile + Hinter.PostRun (forge M11) for the shared
  update-hint wiring, alongside WhatsNewCommand for the whatsnew command.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

Commit Statistics

• 3 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 3 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.

v0.26.0

[0.26.0] - 2026-06-10

🚀 Features

• (app) Propagate remote_metadata policy + report cliff degrade - (a465871) by @bchatard

  withEnvDerivations now carries the top-level remote_metadata policy onto the
  changelog/release-notes drivers so the pipeline generators honour it. CheckCliff
  gains the policy (applied to a driver copy, never mutating the caller's) and
  returns whether it degraded to --offline; heraut check surfaces that in the
  cliff detail line (valid → valid (offline — remote metadata unavailable)).
  Part of T78.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (cmd) Add --offline flag forcing remote_metadata: disabled - (2dffbe0) by @bchatard

  Persistent --offline flag overrides the configured remote_metadata policy to
  disabled for a single run (no remote PR/MR metadata API calls). Wired into
  release, changelog, and the check / check cliff commands via applyOfflineOverride
  after config load. Part of T78.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (config) Add remote_metadata policy field + validation - (b9a467c) by @bchatard

  Top-level remote_metadata governs whether content generators fetch PR/MR
  metadata (author handle, PR number) from the platform API to enrich
  changelog/release-notes: required | optional | disabled, empty resolving to
  optional. Enum-validated. A programmatic ContentDriver.RemoteMetadata field
  carries the effective policy onto each driver for the generator to honour
  (set by the app layer, mirroring HeadingVersionPattern). Part of T78.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (config) Add remote_metadata to schema.json + sample + fixtures - (7aa259f) by @bchatard

  Expose the remote_metadata enum (required | optional | disabled, default
  optional) in schema.json for IDE validation, document it in docs/heraut.sample.yml,
  and add valid + invalid schema fixtures (the invalid one asserts the enum is
  enforced). Keeps the schema/sample/field in sync per the config rules. Part of T78.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (generators/gitcliff) Honor remote_metadata policy via --offline - (e14c020) by @bchatard

  git-cliff fetches PR metadata (and panics on failure) whenever the templates
  reference commit.remote.*. Route Generate and CheckCliff through a policy-aware
  runCliff: 'disabled' always passes --offline; 'required' never does (a remote
  failure stays fatal); 'optional' (default) tries online then retries with
  --offline on any failure, exposing Degraded() so callers can warn. If the
  offline retry also fails the original online error surfaces, so a genuine
  config error still bubbles up. Part of T78.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (pipeline) Surface remote-metadata degrade as a step sub-result - (7c682e1) by @bchatard

  When a generator falls back to --offline because the remote metadata fetch
  failed (remote_metadata: optional), the changelog/release-notes step now carries
  a 'remote metadata unavailable — PR authors/numbers omitted' sub-result, so the
  omission is visible during release/changelog runs. Generators report it via an
  optional Degraded() interface the pipeline type-asserts, keeping the pipeline
  decoupled from the concrete generator. Part of T78.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

📚 Documentation

• (adr) Add ADR-0023 remote-metadata policy; complete roadmap T78 - (fb22f0a) by @bchatard

  Record the remote_metadata decision (top-level tri-state, default optional,
  backed by git-cliff --offline, retry-on-failure for optional, --offline flag) and
  its reconciliation with ADR-0022. Index it in the ADR README and flip T78 to done
  with a completion note.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (adr) Record why the CLI override is boolean --offline (ADR-0023) - (e93cb06) by @bchatard

  Capture the decision to keep --offline as a boolean escape hatch rather than a
  --remote-metadata mirror of the config key (git-cliff precedent + heraut's
  runtime-flag convention + YAGNI), so it is not re-litigated later.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

• (roadmap) Add T78 — remote_metadata policy for git-cliff enrichment - (9680617) by @bchatard

  Captures the durable opt-out for git-cliff's remote-metadata fetch: a
  top-level tri-state policy (required | optional | disabled, default optional)
  governing both changelog and release-notes generation, backed by git-cliff's
  --offline flag. The optional default degrades + warns on fetch failure, fixing
  the tokenless-panic residual out of the box. Follows the ci: GITHUB_TOKEN patch
  that authenticates the fetch in CI.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

⚙️ Miscellaneous Tasks

• Authenticate git-cliff GitHub API calls in release workflow - (cc8f128) by @bchatard

  heraut's changelog/release-notes generation shells out to git-cliff, which
  fetches PR metadata from the GitHub API to populate commit.remote.username and
  commit.remote.pr_number. git-cliff reads GITHUB_TOKEN, not GH_TOKEN (which only
  gh/glab consume), so the Preflight and Release steps were calling the API
  unauthenticated — hitting GitHub's 60/hr shared-IP rate limit and panicking with
  a 403 (exit 101). Export GITHUB_TOKEN to both steps so the calls authenticate
  against the 1000/hr limit.

  Co-Authored-By:Claude Opus 4.8 noreply@anthropic.com

Commit Statistics

• 10 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 10 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.

v0.25.1

[0.25.1] - 2026-06-10

🐛 Bug Fixes

• (deps) Run go mod tidy - (bc8a7b9) by @bchatard

💼 Other

• (deps) Bump forge to v0.11.1 - (82dd26c) by @bchatard

  Tracks the latest forge tag (carries forge's own golang.org/x/sys
  v0.45.0 CVE bump; heraut already pinned x/sys v0.45.0 directly, so no
  effective change there). go.mod/go.sum only.

  Co-Authored-By:Claude Sonnet 4.6 noreply@anthropic.com

🚜 Refactor

• Align forge/log import alias to forgelog - (14750f4) by @bchatard

  Rename the forge/log alias flog → forgelog so it matches the family's
  forge convention (forgeui, forgeexec) and the alias bifrost uses.
  No behaviour change.

  Co-Authored-By:Claude Sonnet 4.6 noreply@anthropic.com

Commit Statistics

• 3 commit(s) contributed to the release.
• 0 day(s) passed between the first and last commit.
• 3 commit(s) parsed as conventional.
• 0 linked issue(s) detected in commits.
• 1 day(s) passed between releases.