Repository navigation
Releases
v0.5.0
Minor Changes
-
#44
67c41a6Thanks @Paalt! - The review flow now detects mistake classes and hardens them instead of patching them again, and the feature-map join stops over-selecting journeys.fix-the-classjoins the toolkit (adapted from pstack'scorrect): after a correction the repo has seen before, or as a hardening pass, find the mistake classes from the repo's own record and make each impossible at the highest level that works — architecture, types, a check whose error names the fix, a test, prose last — proven red on a real past instance, with anAGENTS.mdtable of rules and what enforces them.review-prmarks classes mechanically (Phase 4, after the arbiter): a finding getsclassEvidencewhen it has several sites in the diff, cites a rule the rules table marksprose onlyor does not list, or matches a defect an archived review or an openreview-debtissue already recorded. The report gains aclasschip and a Classes card; the published review body lists the classes before the sections.resolve-reviewhardens class findings instead of only patching them: every site fixed, thenfix-the-classat the guard levels (types, check, test) proven red on the instance just fixed, a rules-table row added or raised, architecture moves handed back as their own PR, and every sibling left outside the branch's scope named byfile:line. The ledger gains a Classes block.- The feature-map join narrows by symbol. A changed file listed by several features keeps a feature only when another of its changed Paths matched directly or one of its Paths references an export the diff changed; a file with no exports drives every feature that lists it (field: a three-file refactor of a shared helper selected eight journeys). Features set aside are recorded as
not-affected; the Driven-features block gains that word. - Launch works from a fresh worktree.
create-verify-skillrequires Launch to name what a worktree lacks (gitignored env files, per-checkout deployments, caches) and the repo's own setup command that provides them; refresh re-audits it. - Contract tests pin the
classEvidencefield name, the rules-table heading, and the narrowing phrase across producers and consumers.
-
#47
2f92589Thanks @Paalt! - The plugins are renamedhp-review-toolkitandhp-code-publish, and their home is now humanityplus.solutions/commons/plugins. The Claude directory listing thehumanity-plus-*names were made for isn't happening, so the plugins are published through this repo's marketplace. If you installed either plugin, swap it:claude plugin uninstall humanity-plus-review@humanity-plus claude plugin marketplace update humanity-plus claude plugin install hp-review-toolkit@humanity-plus
Do the same for
humanity-plus-publish@humanity-plus→hp-code-publish@humanity-plus. If you are still on the names before those (review-toolkit@humanity-plus,code-publish@humanity-plus), uninstall those instead. Add--scope project(orlocal) to the uninstall if you installed there. If your settings list the plugins underenabledPlugins, rename those keys. Slash commands move too, for example/hp-review-toolkit:review-pr. The skills CLI lane is unchanged.The Review Toolkit's full guide moves from
review-toolkit.htmlin this repo to its own page, and the READMEs link there. The toolkit README's "What it runs and sends" now covers the verify skill, which reviews and fixes on your own checkouts use to start your app, andinterrogate, whose external seat (today Codex) sends its brief to that provider under your own login. The bundle READMEs go back to the two-step marketplace install.
Patch Changes
-
#46
0e81d10Thanks @Paalt! - Gate selection proves coverage, and every lens has a materials recipe.- Pick gates from CI, and prove each one covers the changed files. The baseline and the fix phase take the gate set from the CI workflow, by name, never the nearest-sounding package script, and record which changed files each gate reaches — the tool's globs and the workflow's own path filters and job conditions, so a job that did not run covers nothing (field: a workspace
lintpassed while CI'slint:cicaught script formatting it never scanned). A green CI run still stands in for running the gates, never for that pairing; a changed file no gate covers is recorded as a hole in the review's limitations and the fixer's over-to-you list. Stated once inreview-pr;resolve-reviewandreview-fixcite it, and a contract test keeps the three in step and refuses any return to script-name discovery. - Per-lens materials recipe.
review-prstates in one table what every lens, and a generalist carrying several lenses, is handed beyond the common set; the context bundle is fenced by section so a doc-heavy bundle costs the bug lenses nothing and a small review spawns mechanically. - Contract phrase checks match on whitespace-collapsed text; the strings an agent copies into a shell still must sit on one line.
- Pick gates from CI, and prove each one covers the changed files. The baseline and the fix phase take the gate set from the CI workflow, by name, never the nearest-sounding package script, and record which changed files each gate reaches — the tool's globs and the workflow's own path filters and job conditions, so a job that did not run covers nothing (field: a workspace
v0.4.0
Minor Changes
-
#38
1bb2a68Thanks @Paalt! - Newblast-radiuslens inreview-pr, for changes that ship to production. It names the one safety fact the change rests on, proves it as far as reading allows on an evidence ladder (asserted → documented → traced → command-settled → observed, the same names asverifiedBy), writes the probe that settles it by execution, walks the change's reach past grep (other readers of the same bytes, the library at its pinned version, the path production actually runs, timing, rollback), and reports what it cleared as first-class output. On your own checkout the orchestrator runs the probe, so the fact arrives proven or contradicted. The report leads with a Blast radius card,publish.tsputs the fact at the top of the GitHub review body, andresolve-reviewre-runs the probe after its fixes. Runs from the medium tier up; trivial and light diffs skip it. The method adapts theblast-radiusskill in Lauren Tan's (poteto) MIT-licensed pstack, credited in the toolkit README. -
#41
b5389a5Thanks @Paalt! - Newcreate-verify-skillskill: generates — or refreshes — a repo-localverify-<app>skill that drives the real app the way a user does (launch, doctor, personas, drive, evidence, cleanup) plus a feature map with one file per user-facing feature, its source Paths and the seeded Personas it needs, proven by an observable end state. The generator interviews the repo, writes the skill into the host's project skill directory, and drives one feature end to end before handing over; refresh mode runs source readers per feature and one live pass and leaves only proven corrections in the working tree (a PR when asked). Both modes run on your own checkout against a local or disposable instance. The toolkit consumes the skill by shape:review-prruns experiment claims and the blast-radius probe through its recipes on owned checkouts, andresolve-reviewdrives the features a review touched after its fixes land, with the outcome in the resolution ledger. The method adapts thecreate-verification-skillandmaintain-verification-skillskills in Lauren Tan's (poteto) MIT-licensed pstack, credited in the toolkit README. -
#40
fef7ba8Thanks @Paalt! - Newinterrogate[=N]arg forreview-pr: the pipeline's one offensive pass. N adversaries (default 2), each in its own scratch worktree, are briefed to break the change by running it — three focuses, inputs and boundaries / state and time / environment and integration, assigned round-robin — and a finding survives only when the orchestrator's rerun in a fresh worktree reproduces it; the repro command, patch, observed and expected output travel with the finding and render on its card. Adversaries execute code, so the arg is honored only on an owned checkout or underREVIEW.md's## Trust. Runners are detected, never assumed: seats run on the host's subagents, and when the Codex CLI is onPATHone seat runs there so the adversaries span model families; with one family the report says diversity came from the briefs. An Interrogation card shows each seat's found / kept / dropped ledger and the interrogate-only findings. Off by default, likeholistic. The method adapts theinterrogateandarenaskills in Lauren Tan's (poteto) MIT-licensed pstack, credited in the toolkit README.
Patch Changes
- #43
5dcaca0Thanks @Paalt! - Field round from a five-PR queue run.review-pr: a repeated mechanical substitution (the same literal, rename, or import swap in every hunk) takes the light tier however many files it touches, and the light-tier text now defers to the ladder row instead of restating it; when the conflict probe hits most of a PR's files and base replaced what the PR edits, the review asks whether the PR is superseded — goal met on base by reading and measurement — before planning any merge. Theintentlens compares equivalence and duplication claims (keys, resolved values, exports) instead of accepting them.recap.tsemits a one-line classification-only block when no mapped primitive was touched, instead of an empty diagram and table.reset-workspace.shleavesreview-report/verify/in place when it archives stale findings — evidence dirs are keyed by phase and SHA, and moving them dangled every link the findings recorded. A running verify instance and the review scripts no longer collide: the generator now mandates runtime state (pids, port, head, log, registered clients) in the git dir —$(git rev-parse --git-dir)/verify/, shared registrations one level up in the common dir — soreview-report/verify/holds evidence only;check-clean.shrefuses while those pids are alive and fails loudly, leaving the original where it names, instead of nestingreview-report/review-reportwhen a launch recreated the directory mid-check; a contract test pins the pid-file path across generator and script. Existing generated skills pick the layout up on their next refresh.create-verify-skill: identifiers a generated Launch derives from the checkout path are sanitized to the target's charset, Doctor prints them, and refresh mode re-reads Launch, Doctor, and Cleanup against the generator's contract so existing skills pick the rule up. Review-fix on this round added: a superseded verdict is alwayscomment, neverapprove; the review-pr README defers to the ladder row and a contract test keeps the light-tier predicate stated once; the light generalist owes a substituted token the hidden-reader grep that blast-radius would have done at a higher tier.
v0.3.5
Patch Changes
- #36
74a45c6Thanks @Paalt! - Both plugins now link their directory listing to a setup guide at humanityplus.solutions/commons/plugins, which covers theREVIEW.md, convention docs, and primitives map the Review Toolkit reads, and the repo files Code Publish checks. The listings now declare their icons and link to support (GitHub issues) and a privacy notice stating that the plugins collect nothing. The plugin READMEs point to the same guide, and the Review Toolkit's setup list now includes the convention docs step it was missing.
v0.3.4
v0.3.3
v0.3.2
Patch Changes
-
#29
1347b5fThanks @Paalt! - Both plugins now ship an icon: the Humanity Plus leaf. The review toolkit guide links to Bun's install page instead of showing a download-and-run command, and both READMEs state that the skills never read a token or key. -
#31
7e31583Thanks @Paalt! -humanity-plus-publishgets its own icon, the Humanity Plus leaf with a publish badge, so the two plugins are easy to tell apart in the directory.
v0.3.1
Patch Changes
- #27
945391eThanks @Paalt! - Three lens definitions reworded in 0.3.0 had quietly narrowed what they check; they're restored to full breadth in our own phrasing.zombiesBoundaries covers time edges again (expiry at the boundary, clock rollover, DST — not only timeouts), Many covers concurrency again (async interleaving and ordering races — not only parallel access), andtriage's always-high rule covers any authentication (API keys and service-to-service auth included — not only user sign-in and sessions).
v0.3.0
Minor Changes
-
#24
3d80f7fThanks @Paalt! - The plugins are renamed for the Claude directory:review-toolkitis nowhumanity-plus-reviewandcode-publishis nowhumanity-plus-publish. The old names stop resolving once you update the marketplace, so if you installed either plugin, swap it for the new name:claude plugin uninstall review-toolkit@humanity-plus claude plugin install humanity-plus-review@humanity-plus
Do the same for
code-publish@humanity-plus→humanity-plus-publish@humanity-plus, and add--scope project(orlocal) to the uninstall if you installed there. If your settings list the plugins underenabledPlugins, rename those keys rather than adding new ones. Slash commands move with the plugin, for example/humanity-plus-review:review-pr. The skills CLI lane is unchanged.Each bundle README now has a one-step install line, says the skills are built for Claude Code, and lists everything the skills run and send. A few lens instructions in
review-pr(triage, preflight, first-five, zombies) are reworded without changing what they check.
Patch Changes
-
#26
6250553Thanks @Paalt! -review-prandresolve-reviewfixes, from a CodeRabbit review of a repo that vendored the skills:- A
findings.jsonfrom an earlier commit no longer survives new commits on the same PR.reset-workspace.shkeeps it only when it describes the checked-out commit, andresolve-reviewstops and asks before editing from findings about different code (#19). - Finding text can no longer close the report's own containers to render a spoofed verdict outside its card (#20).
- Copying
REVIEW.example.mdno longer opts you into running colleagues' PRs: theexecution: collaboratorsline ships commented out (#21). If you copied the template before this release, check yourREVIEW.md's## Trustsection. check-clean.shno longer runs the check when it can't movereview-report/aside (#22).SKILL.mdgives one rule for when theseamslens runs: whenever triage finds interfaces between areas (#23).
- A
v0.2.0
Minor Changes
- #15
8cfe5f2Thanks @Paalt! - Team/org reviews: repos can opt colleagues' PRs into execution trust (## Trustin REVIEW.md,execution: collaborators— collaborator-authored, same-repo-head PRs get baseline checks, experiments, and merged-tree gates in a disposable worktree), the seams lens traces migrated grants through every authorization consumer, skeptic verification can reframe a finding to the lesser defect that survives instead of discarding it, and resolve-review states reviewed-SHA vs current head and recommends a re-review after substantial security fixes. Publishing on private repos:publish.ts --pages owner/namehosts the HTML report via a Pages-publishing repo (Enterprise access-controlled Pages) and--recapputs the system-recap block in the review body for PRs whose description you don't own; the review footer timestamp is now human-readable (DD.MM.YYYY - HH:MM). create-pr includes the system-recap block at PR creation when a primitives map exists, and gains a retrofit mode that rewrites an existing PR's title and body to the same standard.
Patch Changes
- #16
b9e1e8fThanks @Paalt! - Report fix: the Dependencies (WARM) table no longer overflows its card. Long unbreakable content in the Notes column — 40-char commit SHAs, command strings in inline code, embedded markdown tables — used to set the column width, pushing the table past the card edge and crushing the Package column to one word per line. The table now uses fixed layout with proportional columns, and inline code and embedded tables wrap inside the Notes cell.
v0.1.1
Patch Changes
-
#12
6e983d7Thanks @Paalt! - Midnight redesign (#11): the toolkit guide and the review report now share one visual language — near-black indigo-tinted surfaces, hairline borders, Inter + JetBrains Mono. The report renderer (render.ts,recap.ts) and the finding-style lens were updated to match, so reviews produced byreview-prpick up the new look. -
#9
4b113c7Thanks @Paalt! - Add a unified release pipeline with changesets. The version inpackage.jsonis now the repo's release version;scripts/sync-plugin-version.mjsstamps it into everyskills/*/.claude-plugin/plugin.json, the release workflow maintains a version PR and tags a GitHub Release on merge, and CI fails PRs whose plugin versions drift out of sync. The README documents how each install lane receives updates.