Version 0.3.0

Changelog

• 19d9685 fix(lint): drop nolint directives gosec no longer needs
• 0752195 build(release): point homebrew cask at hackmajoris/homebrew-apps
• b80ae86 Merge pull request #1 from hackmajoris/feat/opencode-support
• 319c81a feat(opencode): added opencode support
• 33ede7a build(deps): bump github.com/stretchr/testify from 1.11.1 to 1.12.1
• 4b87635 fix(tests): end the rollout liveness case on the tail's own touch
• 64b4fe3 fix(tests): use strings.Cut where only the first field is read
• 0ac2919 docs: update changelog for v0.2.0
• e01f8ac fix(tests): exit the fake CLI through syscall.Exit
• 52052d6 chore: store vendored files without eol conversion
• d435850 build(deps): bump github.com/charmbracelet/x/ansi from 0.11.7 to 0.11.8
• 72263b5 fix(prompts): name the adversarial agent for its lens, not its binary
• af0b360 resolve a project-level profile.md under the round's own (umputun#12)
• 9f33422 fix: build the version timestamp with git instead of BSD date
• bf92e03 docs: backlog the immaterial gloss and the lens-override panel body
• fb79350 fix: name the entry when an escaping link occupies a directory name
• 876fc82 docs: align go comments with the comment rules
• 736053b feat(prompts): flag redundant code comments
• 6bbe2f7 docs: point the hero at the skill install
• 0f16f55 docs: revert the hero to a single install command
• 0cb124f docs: put the skill install beside the binary install
• 774cd46 fix(ui): stop closing the tty out from under bubbletea's resize check
• e3765e2 docs: drop the go report card badge
• 96010bd docs: add build badges and a site nav bar to the README
• 6c1051f docs: show the TUI in the README
• 0cb57ff docs: changelog for v0.1.0 and plugin 0.4.2
• 7d741d3 docs: add revmux.com and trim the README to a synopsis (umputun#6)
• c4a8c3c build: put shellcheck behind make lint
• 413321b fix(skill): write the session-field guard so shellcheck stays quiet
• 9e14a94 fix(skill): recover missing agterm context
• fe0cd66 fix(executor): feed the idle watchdog through claude's structured-output emit
• 7377a5d fix(skill): stop the loop on a repeated finding, not on a flat count
• 179d596 docs: name the effort and the real models in every expert row
• c3f62de feat(prompt): add the expert profile, two top models at the highest effort
• 8602322 docs(skill): put the shape of the presented turn in one reference
• 8411b59 fix(skill): frame a review — what it is, what it amounts to, and why
• 5e06ead feat(report): record what each source raised, so a blind agent is visible
• 1c40407 fix(skill): make a triage end on a verdict and a question, not a wall of arguments
• e7825a0 feat(ui): drop the input tab for an optional file nobody supplied
• 1eb59c9 fix(ui): wait for bubbletea before closing the tty
• 6b58333 docs(lens): tell every triage lens to report an inability to work
• 72fcde1 add issue triage, a four-way panel over a filed item (umputun#4)
• e7633eb fix: sweep the minors from the since-8e8d116 review round
• 83cd960 docs(skill): budget the round-preparation brief, and bump the plugin to 0.2.8
• de9a63b feat(ui): name the profile in the header instead of counting agents
• 8e8d116 feat(prompt): add the grill-me profile, four adversarial agents over two vendors
• 6b841f5 fix(skill): tint the floating overlay too, not only the pane one
• 90d259b refactor: cut comment volume to sibling levels [skip ci]
• f2719f1 fix(archive): make every function in archive.go spell dir, path
• 2078b40 feat: measure what the review stages do, and ship the corpus analysis as a script
• cd56094 feat(cleanup): add revmux cleanup, and the archive numbers it is decided from
• b3f840b fix(ui): drop the input tab for an absent or empty context
• 8104eb9 fix(skill): keep the progress feed out of an overlay run, and route "show me" to it
• ed79c1a fix(ui): join a tight definition description's soft break too
• e7392cb fix(ui): stop joinTightListLines corrupting blockquotes, and make it see CRLF
• 1eb1d4c feat(ui): move to glamour v2 and reflow soft line breaks in list items
• bfed100 feat(skill): use a pane overlay when the launcher runs in an agterm split
• 20adae4 feat(skill): relay run milestones while a headless review runs
• a3a59f0 test(ui): pin ordered list marker spacing, record the missing hanging indent
• e84591e fix(ui): drop glamour's padding around an inline code span
• 203d215 chore: bump dependencies and revendor
• ea14e79 fix(ui): scope document cache eviction to the pane layout pass
• b5c28b2 fix(executor): drop shell tool calls carrying no command
• d753bd9 docs: move completed plan 20260731-glamour-markdown-rendering.md to completed/
• 5e3dcd6 docs: rewrite the tui rules and README for the new rendering and keys
• 45bf072 feat(ui): stop esc and q from quitting a running review
• 414f693 feat(ui): render markdown documents with glamour
• eaa425d chore: add glamour dependency and vendor its tree
• 48b07a2 docs: add glamour markdown rendering implementation plan
• 4cc0a20 fix(skill): rewrite herdr tab cleanup as an if for shellcheck
• 09904c1 docs: add CONTRIBUTING and CHANGELOG
• b41a376 chore: add goreleaser release pipeline, dependabot, funding and codeowners
• d8902a1 docs: explain mixed-model review profiles
• 1dfa345 fix(skill): include severity in finding tags
• 1820e21 fix(tui): harden startup input viewer
• 8498085 docs(skill): tag every reported finding with the surface it sits on
• 27f0475 chore(plugin): bump version to 0.2.4
• bf8aedf fix(prompts): tie severity to runtime consequence, not to how wrong a claim is
• 6a6ec3a feat(tui): add startup input viewer
• adae841 feat(prompt): one model string selects the runner, for the whole review
• 1be1942 chore(plugin): bump version to 0.2.3
• bd8b348 docs(skill): add a PR target to the revmux skill
• 5d2ea89 chore(plugin): bump version to 0.2.2
• 803054f docs(skill): add loop mode to the revmux skill
• deda9ca docs(skill): pick the re-review profile from what the fixes touched
• 56c476b fix(ui): color the header findings count by worst severity
• 83eedee feat(prompts): add claude-only and codex-only profiles
• 918123c fix(prompts): never drop a critical or major for lack of corroboration
• 374de5e feat(prompts): tell review agents not to run tests, builds or the linter
• 982c9ac fix: keep looking for the codex rollout instead of globbing once
• ab16c7a fix: contain the init config write in the destination root
• d3489e4 docs: move completed plan 20260728-init-and-self-reflection.md to completed/
• 7e6025a fix: address code review findings
• d54a062 fix: make the dangling-symlink guard test reach the code it guards
• 1fa3ac9 fix: address code review findings
• d97c0e2 docs: document revmux init and revmux stats
• 5a8a89c docs: record task 10 acceptance verification results
• 6c4a072 docs: add the revmux self mode to both skill trees
• cc06127 feat: add the revmux stats subcommand
• 2383980 feat: aggregate rounds and tasks into a corpus
• 0852e4b feat: read one round's review numbers in app/archive
• 3ff20b2 feat: fold --init into the init subcommand
• bd92281 feat: add the revmux init subcommand
• ffef1e1 feat: retain and expose the winning prompt file's bytes
• 9e952df refactor: lift task enumeration into app/task
• f7350c2 refactor: join every archive path from app/task
• 6bb8491 docs: add plan for revmux init, revmux stats and /revmux self
• 05d9b7b test: make the width cases discriminate, and close the emacs sentinel race
• cd38ad6 fix: stop the ERR trap discarding a review it was added to protect
• 6d554da fix: keep launcher failures out of revmux's exit-code vocabulary
• 95bb60f fix: stop claiming a run completed when its archive did not
• 4f9294f feat(skills): close a review with what is actionable and what to do next
• 5d2ac1b feat(skills): resolve a profile word onto a profile name
• 81f3bf7 feat(prompt): weigh a fix's blast radius against the finding's severity
• 90dbc17 feat: close the stderr log with what the run came to
• 9a5326c fix: correct claims the layout change left stale
• e6dd65f refactor: drop old-layout task detection
• 5369dc5 docs: mark the plan complete and move it to completed/
• dafc355 docs(plan): fold the coupled layout tasks together and mark the plan done
• 27b0ef8 build: check the two skill trees agree
• de76d56 docs(skills): drive revmux new instead of constructing paths
• 0bb8476 docs: describe the round layout, and record what it removed
• 9fb317e feat: add revmux new, and report the task store in revmux config
• 1bfee57 feat: move the review context into the round, and stop deleting
• b12cd2f feat(task): add app/task for the round layout and its name checks
• 4e6324a refactor(prompt): move front-matter splitting into app/frontmatter
• 1f21e93 docs: drop pruning from the task/round layout plan
• cf23fa7 docs: expand readme with the reasons revmux exists
• 08f386c docs: add plan for moving review context into the round
• d97be84 docs(skills): cut rationale from the skill, add task and run naming rules
• d095e9a fix(skills): correct launcher exit codes, unbounded waits and task-id validation
• 4ed3259 feat: ship the caller as a skill for claude code and codex, add make install
• 1c6caf4 fix(ui): render a title before wrapping it, not after
• 6554ba5 fix(ui): bound every row Wrap returns, and style a wrapped title per row
• 2bb83b2 refactor(ui): one Wrap instead of three, and stop spans turning headings off
• 94e1b0b fix(ui): render markdown in every pane, and rebuild the header count from the report
• 995f637 feat: put JSON on stdout by default, markdown behind --markdown
• ce8a9e9 feat(ui): break the header's findings count down by severity
• c06c558 fix(ui): stop the clock when an agent stops, and render the report as a report
• 6ff7cdb fix(ui): keep every row's clock running, and separate the header from the table
• d232f74 fix(ui): collapse the fewest tabs that fit, and the padding before any name
• 36600a8 docs: record what the rework changed, and what it cost to learn
• 4c75777 feat: add --auto-exit, and give the plain renderer the width it was missing
• 576f483 feat(ui): rework the frame, the findings browser and the tab bar
• 9ee5076 feat(executor,pipeline): supervise codex by its rollout, and report what agents are doing
• 908726a feat(prompt): color stage processes, and state the read-only rule in terms of shells
• 8d1dd9e fix(archive): make run pruning correct on linux, not just on macos
• 9b3e9c1 chore: vendor golang.org/x/term
• 7a082c8 fix: address code review findings
• 0753fcb feat: update documentation
• d459fa5 feat: verify acceptance criteria
• ce46766 feat: add revmux config introspection command
• 5f44d01 feat: add full default prompt set
• a8cbda1 feat: add TUI findings browser
• 3dcf817 feat: add TUI status table, combined view and per-agent panes
• 4b7eb47 feat: add run artifacts
• 7fac55c feat: add verification stage
• 3198da5 feat: add synthesis stage
• 60c6565 feat: add codex executor as a peer source
• 18a3651 feat: add parallel fan-out, staggered launch and degrade policy
• 07138f1 feat: add find stage, event channel and end-to-end slice
• 65cb0ba feat: add configuration, CLI options and task directory resolution
• 9613afa feat: add prompt loading, roster parsing and lens composition
• f9b60e7 feat: add supervised claude executor
• e542a03 feat: add finding types, report rendering and stage schemas
• 6473b3f feat: bootstrap go project with makefile, linter config and CI
• 5d0de7f docs: apply review findings to rules and build plan
• 37f26f4 docs: add initial build plan
• 7340365 docs: add project notes and path-scoped subsystem rules
• 9a4ca67 Initial commit