v0.23.0

0.23.0 (2026-09-04)

Features

• rook-triage: bound the issue mine and ship the label-list check (77b1ed7), closes #97
• rook-triage: deep-fetch truncated PRs before any flag is resolved (03824b3), closes #96

Bug Fixes

• rook-triage: assign identity resolution to a stage that exists (8f609f5), closes #94 #95 #98
• rook-triage: state four drifted rules once, at their homes (7dba576), closes #99

What's Changed

• feat(rook-triage): rebuild the kb refresh on the shipped tools by @jhoblitt in #113

Resolved issues

• #96 kb refresh: deep-fetch before truncation flags reach the resolver
• #97 kb refresh: the issue-participation and label-list sources are unbounded and untooled
• #98 kb-refresh.md assigns identity resolution to a miner that no longer exists
• #99 rook-triage: four stale or duplicated renderings left after the kb-refresh split

v0.22.0

0.22.0 (2026-09-04)

Features

• rook-triage: check the kb against prev, CODE-OWNERS and the fetch state (dc2eb07), closes #94
• rook-triage: emit the CODE-OWNERS roster from rt-analyze (bdd67a0), closes #95
• rook-triage: ship the issue mine as rt-issues (e1e0ab9), closes #97
• rook-triage: ship the label-map diff as a validate-actions mode (bf8d5d2), closes #97

Bug Fixes

• rook-triage: bound and fence the text rt-analyze mines (5b56a4f), closes #101
• rook-triage: correct the classifier's documentation and its zero-match tests (76e5df4), closes #102

What's Changed

• feat(rook-triage): ship the kb refresh's gates and mines as tools by @jhoblitt in #112

Resolved issues

• #94 validate-kb: move the assembler's remaining deterministic checks into the tool
• #95 kb refresh: derive roster from CODE-OWNERS in code and drop the source-1 miner
• #101 rt-analyze and rt-commits emit contributor-authored strings unfenced
• #102 rt-analyze: documentation and test nits left by the classifier fixes

v0.21.0

0.21.0 (2026-09-03)

Features

• rook-code-review: add cross-cluster coupling to the race surface (e2239ae), closes #103

Bug Fixes

• rook-code-review: reduce drifted renderings to their one home (83d367c), closes #105
• rook-maintainer: pin the Bash grant and derive the routed set (186122c), closes #106

What's Changed

• test(rook-maintainer): pin the four rules that moved in v0.19–v0.20 by @jhoblitt in #110
• fix(rook-code-review): cross-cluster coupling, one-home renderings, derived routed set by @jhoblitt in #111

Resolved issues

• #103 rook-code-review: the gate's failure-surface list never names cross-cluster coupling
• #104 evals: pin the four review rules that moved or landed in v0.19–v0.20
• #105 rook-code-review: drift leftovers the gate noted outside the #90/#93 diffs
• #106 rook-reviewer: pin the Bash-grant rationale and derive the expected routed set

v0.20.1

0.20.1 (2026-09-03)

Bug Fixes

• rook-conventions: fill four canon gaps in fan-out and PR rules (362f43d), closes #107

What's Changed

• fix(rook-conventions): fill four canon gaps in fan-out and PR rules by @jhoblitt in #109

Resolved issues

• #107 rook-conventions: four canon gaps the gate noted

v0.20.0

0.20.0 (2026-09-03)

Features

• rook-conventions: check backported workflows against the branch (bf7e596), closes rook/rook#18232 rook/rook#18312 #89
• rook-conventions: derive backport set from Ceph version support (55ca7b7), closes rook/rook#18242 #78
• rook-conventions: keep a blessed PR's backport labels current (a4aa870), closes rook/rook#18242 #79
• rook-conventions: verify non-owner technical claims before acting (69a47f3), closes rook/rook#18242 #76

Bug Fixes

• rook-conventions: keep PR title current, and on additive pushes (a1c06e4), closes rook/rook#18218 #72
• rook-conventions: route reviewer selection via rook-triage's KB (d4cda26), closes rook/rook#18242 #75
• rook-conventions: say how a status-check name is composed (8baac80), closes rook/rook#18232 #74

Refactoring

• rook-conventions: split label application out of backporting.md (2ec7202)

What's Changed

• feat(rook-conventions): close seven house-rule gaps from recent rook PRs by @jhoblitt in #93

Resolved issues

• #72 rook-conventions: PR-update rule covers the description but not the title
• #74 rook-conventions: workflows-and-ci.md never says how a status-check name is composed
• #75 Reviewer selection must use the rook-triage routing KB, not ad-hoc CODE-OWNERS
• #76 Treat PR-comment analysis/direction/implementation skeptically when the author isn't in CODE-OWNERS
• #78 Backport eligibility for version-specific fixes should use a version-support × affected-range matrix, re-evaluated on range changes
• #79 Once a PR is blessed for backporting (CODE-OWNER or existing label), keep backport labels in sync as the PR changes
• #89 rook-conventions: backporting.md guards docs against master-only refs but not workflows

v0.19.2

0.19.2 (2026-09-03)

Bug Fixes

• rook-maintainer: ignore another plugin's CLAUDE_PLUGIN_DATA (e149d12), closes #85
• rook-triage: mine origin/master in rt-commits --repo (de49593), closes #84
• rook-triage: ship real logins in the kb snapshot (da812bf), closes #83
• rook-triage: stop excluding travisn from routine routing (5c0c38f), closes rook/rook#18242 #77

Refactoring

• rook-triage: split kb refresh out of routing.md (b10a921)

What's Changed

• fix(rook-triage): correct the routing KB and its mining tools by @jhoblitt in #92

Resolved issues

• #77 routing-overrides.md '*: travisn — escalation-only' is incorrect; travisn is a routine reviewer
• #83 kb-snapshot.json ships two display names in the maintainers login field
• #84 rt-commits --repo mines HEAD, but routing.md specifies origin/master
• #85 run.sh trusts an ambient CLAUDE_PLUGIN_DATA that may belong to another plugin

v0.19.1

0.19.1 (2026-09-03)

Bug Fixes

• rook-triage: anchor the ceph-external and block substring rules (b142f81), closes #81
• rook-triage: stamp nested go module manifests as build (aa20123), closes #82
• rook-triage: stop stamping generated artifacts as helm/docs (d702ea0), closes #80

What's Changed

• fix(rook-triage): anchor rt-analyze's area classifier to its spec by @jhoblitt in #91

Resolved issues

• #80 rtanalyze: generated CRD artifacts stamp helm and docs, inflating reviewer authority
• #81 rtanalyze: strings.Contains over-stamps ceph-external (externalversions) and block (CSI rbd templates)
• #82 rtanalyze: pkg/apis/go.mod stamps crd instead of build (exact-equality path test)

v0.19.0

0.19.0 (2026-09-03)

Features

• rook-code-review: hunt cross-cluster reconcile serialization (2c5ca40), closes rook/rook#18241 #73

Bug Fixes

• rook-code-review: make the routed reference set a gap-sweep floor (d64df83), closes rook/rook#18058 #88
• rook-code-review: re-derive findings adopted from review threads (a1221ba), closes rook/rook#18058 #76 #87 #76
• rook-code-review: register the GH Actions action-download flake (dd971ba), closes rook/rook#17952 #86

What's Changed

• ci: pin the shellcheck the validate job runs by @jhoblitt in #71
• fix(rook-code-review): re-derive adopted findings and floor the gap sweep by @jhoblitt in #90

Resolved issues

• #73 rook-code-review: add an axis for CR-reconcile serialization (keep CephClusters parallel)
• #86 known-flakes.md has no entry for GitHub Actions action-download failures, which fail before any repo code runs
• #87 rook-code-review: findings adopted from existing review threads bypass verification and inherit the commenter's severity
• #88 rook-code-review: an orchestrator's focus list can silently narrow a gap sweep, and the empty result is then counted as coverage evidence

v0.18.0

0.18.0 (2026-08-18)

Features

• rook-maintainer: guard the fallback agent inside a rook checkout (5193e15), closes #55

Bug Fixes

• rook-code-review: deny the review spine its in-place writer (4154fa7), closes #60
• rook-code-review: fence the untrusted spans the inline spine reads (10bc0ce), closes #59
• rook-code-review: pin authority docs on every path that reads them (d077514), closes #56
• rook-code-review: stop the inline accuracy pass fetching unguarded (1b3aa89), closes #54
• rook-maintainer: cut a truncated string on a rune boundary (0f78891)
• rook-maintainer: escape the path validate-anchors prints (a3d5813), closes #58
• rook-maintainer: fence the ref names the rebase notice injects (9ef79ea), closes #57
• rook-maintainer: point the deny message at a tool that exists (41f789b)
• rook-maintainer: surface a fetch the guard did not adjudicate (7aed828), closes #54
• rook-maintainer: verify the cached tool binary before exec (49db402), closes #53
• rook-maintainer: warn when the default branch is behind (1b43e8d)

Refactoring

• rook-maintainer: give PR takeover its own skill (47c379d), closes #60

Documentation

• rook-code-review: describe the fallback guard's checkout scope (ec9e014), closes #55
• rook-code-review: name the apiserver audit log as an argv channel (03467ba), closes #69
• rook-maintainer: diagram every procedure skill and keep it current (17fb345)

What's Changed

• fix(rook-maintainer): close the open audit findings by @jhoblitt in #70

Resolved issues

• #53 rook-code-review: tools/run.sh execs a cached binary with no integrity check
• #54 rook-code-review: WebFetch guard does not bind on the inline path, and fails open silently
• #55 rook-code-review: the general-purpose fallback escapes both the agent roster and the WebFetch guard
• #56 rook-code-review: authority-order documents are read from the contributor's checkout
• #57 rook-maintainer: rebase-notice.sh interpolates ref names raw into every prompt
• #58 validate-anchors: path is interpolated unescaped, bypassing the package's own quote()
• #59 rook-code-review: the treat-as-data rule ships without a delimiter
• #60 rook-code-review: the skill declares no allowed-tools and inherits the session roster
• #69 security.md: argv bullet omits the apiserver audit-log channel and can bless a literal env value

v0.17.0

0.17.0 (2026-08-17)

Features

• rook-maintainer: decide PR-template conformance with a tool (0b78116)

Bug Fixes

• rook-maintainer: match dispatch and retrieval to the work (738d981)

Refactoring

• rook-maintainer: give each restated rule a single home (5399787)

What's Changed

• fix(rook-maintainer): close the audit's remaining findings by @jhoblitt in #68