loadout-ai 0.3.1 → 0.4.1

package/CHANGELOG.md

@@ -1,5 +1,72 @@
 # Changelog
 
+## 0.4.1 - 2026-07-20
+
+### Added
+
+- Detect local Node CLI, npm package, release, Vitest, Commander, Zod, MCP,
+  and security signals when recommending project tooling.
+- Label recommendations as skill libraries, explicit MCP/runtime setup, or
+  unavailable instead of presenting every catalog record as automatically
+  activatable.
+
+### Changed
+
+- Calculate project activation capacity separately for each agent and count
+  both Loadout-managed and pre-existing unmanaged skills toward `--limit`.
+- Prefer compact, diverse project skill sets and reject mismatched Jest-only
+  guidance for Vitest-only repositories.
+
+### Fixed
+
+- Allow project activation through recursively empty directories restored by
+  rollback while continuing to block files, symlinks, unreadable paths, and
+  unsupported entries.
+- Re-check per-agent capacity and target occupancy inside the activation
+  transaction so filesystem changes after preview abort without partial edits.
+
+## 0.4.0 - 2026-07-20
+
+### Added
+
+- Add generated, test-enforced README facts and bounded release claims so public
+  documentation cannot silently drift away from the shipped catalog, profiles,
+  adapters, commands, or verification surface.
+- Add disposable filesystem lifecycle conformance coverage for all 12 configured
+  agent adapters and an offline README product journey covering library install,
+  activation, inspection, privacy-safe sharing, and rollback.
+- Add a new Loadout visual identity and a proof-first README organized around the
+  Choose -> Inspect -> Preview -> Apply -> Undo workflow.
+
+### Changed
+
+- Make CI and claim-generation paths portable on Windows, fetch the repository
+  history needed by evidence checks, and pin third-party Actions by immutable commit.
+- Consolidate active planning in `MASTER_PLAN.md` and retain completed implementation
+  documents only as historical evidence.
+
+### Fixed
+
+- Refuse explicit rollback when managed paths changed after a mutation, including
+  dashboard restores and unsupported filesystem entries, while retaining automatic
+  failed-transaction recovery.
+- Bind adoption to the complete safe skill tree, reject drift and forged plans, and
+  record ownership only after verifying the final bytes.
+- Record installed-profile state inside the setup transaction so Stable and other
+  profile changes can be rolled back consistently.
+
+## 0.3.2 - 2026-07-18
+
+### Fixed
+
+- Treat recursively empty skill directories left by an older removal as unoccupied,
+  while continuing to block every directory containing a file, symlink, or other
+  unmanaged entry.
+- Re-check target occupancy immediately before copying or replacing files to close the
+  preview/apply race.
+- Remove recursively empty managed skill directories during complete uninstall so a
+  clean reinstall does not encounter false conflicts.
+
 ## 0.3.1 - 2026-07-18
 
 ### Fixed

package/MASTER_PLAN.md

@@ -6,13 +6,14 @@ Category: Developer Tools
 Team size: 3  
 Target submission: July 21, 2026 at 5:00 PM Pacific / July 22 at 4:00 AM Dubai
 
-## Current remaining work (July 18, 2026)
+## Current status and remaining work (July 20, 2026)
 
 This is the authoritative active list. The long phase history below is retained as
-an engineering record, not a command to build every speculative system before users
-can test Loadout.
+an archival engineering record, not a second active plan and not a command to build
+every speculative system before users can test Loadout. Completed implementation
+plans and contributor-specific plans have been consolidated here and removed.
 
-### Required before calling the CLI polished for public testing
+### Implemented product work
 
 - [x] `P18-01 [TERRA]` Make the first CLI screen beginner-readable: a read-only
       `loadout guide`, concise default library summary, focused first-screen help,
@@ -21,9 +22,6 @@ can test Loadout.
 - [x] `P18-02 [TERRA]` Make core machine output honest: `catalog --json` returns
       JSON, and `update --package <id>` checks only that package rather than every
       tracked installation.
-- [ ] `P18-03 [HUMAN+TERRA]` Run the founder acceptance path in
-      `docs/USER_TEST_GUIDE.md` on the real Codex and Claude profiles. Record each
-      observed failure and turn reproducible ones into regression tests.
 - [x] `P18-04 [TERRA]` Bound and explain live network checks in `health --updates`
       and `watch --once`; they must show progress, a clear timeout, and an actionable
       result even when a large Maximum library is present.
@@ -36,9 +34,169 @@ can test Loadout.
 - [x] `P18-07 [TERRA]` Add a concise beginner section to README that links to the
       testing guide and explains Stable, Power, Maximum, project recommendations,
       daily discovery, and rollback in plain language.
-- [ ] `P18-08 [HUMAN+TERRA]` Run the full verification gate, publish the next npm
-      patch/minor only after founder acceptance, and test that exact tarball in a
-      fresh terminal.
+- [x] `P18-15 [TERRA]` Bind material README claims to checked repository evidence,
+      add generated facts and adapter lifecycle coverage, test the documented product
+      flow, and separate deterministic, live, and human evidence.
+- [x] `P18-16 [TERRA]` Make risky setup previews include every required approval
+      flag, reject unknown top-level commands, and make user-requested rollback fail
+      closed when files changed after a mutation or when a legacy snapshot lacks
+      post-mutation evidence. Internal failed-transaction recovery remains
+      authoritative.
+- [x] `P18-19 [SOL+TERRA]` Re-audit the final unmerged development branch before deletion and preserve its
+      one valuable adoption-safety intent in the current architecture. Adoption now
+      binds the complete safe tree, rejects drift and special entries, records only
+      final verified bytes, conservatively attributes catalog review, deep-freezes
+      previews, and accepts only the exact same-process planner-issued plan
+      (`186daa0`, `09e0e0c`, `3f2cafe`, `10d6109`). No valuable work remains on the
+      retired branch; its local and remote copies were deleted after synchronization while
+      the user's ignored `.superpowers/` directory was preserved.
+- [x] `P18-20 [SOL+TERRA]` Fix the fresh-clone live Stable rollback regression
+      without weakening drift protection. `5f8e38e` records installed-profile state
+      inside the catalog transaction before post-mutation evidence and makes the live
+      evidence journey roll Stable back before unrelated fixture transactions. An
+      unsafe proposal to ignore `state.json` drift was rejected because it could
+      orphan later installations.
+
+All intended work through P18-20 is on remote `main`, and there are no open PRs. The
+merged `codex/relatable-readme-hero` remote branch remains safe to delete after the
+0.4.0 release checkpoint.
+
+### Deterministic repository verification
+
+- [x] `P18-08 [TERRA]` Run the complete clean-state verification gate after plan and
+      documentation consolidation, including formatting, lint, type checking, build,
+      evidence checks, unit and integration tests, CLI and README product flows,
+      package smoke, performance, dashboard Playwright, package contents, and
+      `git diff --check`. The July 20 integrated-main audit passed 113 test files with
+      597 tests passing and one intentionally skipped, both product flows, package
+      smoke, the 1,000-skill performance gate at 2.40 seconds p95, and both dashboard
+      viewports.
+      The additional live Stable gate installed four pinned packages and completed
+      state and filesystem rollback assertions at `5f8e38e`.
+
+### Branch cleanup status
+
+- [x] Review all unique history on the final unmerged development branch before
+      deletion; the adoption gap above was the only remaining valuable intent.
+- [x] Reimplement and regression-test that intent in the current architecture.
+- [x] Complete the second remote synchronization, confirm the four adoption commits
+      are reachable from remote `main`, and delete the superseded development
+      branches. The temporary worktree was removed and the unrelated ignored
+      `.superpowers/` directory was preserved. The later merged README hero branch is
+      retained only until the 0.4.0 release checkpoint.
+
+### Human and external work (not provable by local tests)
+
+- [ ] `P18-03 [HUMAN+TERRA]` Run the founder acceptance path in
+      `docs/USER_TEST_GUIDE.md` on real Codex and Claude profiles. Record each
+      observed failure and turn reproducible ones into regression tests.
+- [ ] `P18-13 [HUMAN+TERRA]` Publish `loadout-ai@0.4.0` to npm, then test that exact
+      registry tarball in a fresh terminal through Stable -> rollback -> Power ->
+      rollback -> Maximum -> project optimization -> complete uninstall.
+      Publication completed on July 20. A clean temporary install resolved the exact
+      registry tarball, reported version `0.4.0`, and completed `loadout demo` with
+      rollback verification. The founder's real Codex/Claude CLI lifecycle test
+      remains required before this item can be checked; dashboard testing was removed
+      after founder review rejected it as a conflicting product surface.
+- [ ] `P18-17 [HUMAN]` Resolve the GitHub account billing or spending-limit condition
+      that prevents Actions jobs from starting, rerun CI and daily discovery on the
+      integrated commit, and record the result. CI runs
+      [`29691581581`](https://github.com/VirajMishra1/loadout/actions/runs/29691581581)
+      (`4fdc473`) and [`29692535521`](https://github.com/VirajMishra1/loadout/actions/runs/29692535521)
+      (`e74ba16`) failed before any step ran; their annotations cite failed recent
+      account payments or a spending limit that must be increased. They are external
+      runner-account failures, not product-test failures.
+- [ ] `P18-18 [HUMAN]` Decide and configure appropriate `main` branch protection.
+      GitHub currently returns 404 for the protection endpoint, so protection is
+      absent or not observable; do not describe it as enabled.
+
+### Founder acceptance findings and 0.4.1 corrections
+
+- [x] `P18-21A [HUMAN+TERRA]` Verify the published 0.4.0 Stable lifecycle on the
+      founder's real Claude Code and Codex paths. Stable installed four pinned sources
+      as 60 managed activations, preserved all 12 unmanaged Claude skills, reported no
+      managed drift, and restored the explicit pre-install snapshot successfully.
+- [ ] `P18-21B [TERRA]` Make rollback history understandable before 0.4.1. The first
+      founder test exposed that bare `loadout rollback` selected a newer no-op
+      dashboard/sync snapshot whose pre-state already contained Stable. No data was
+      lost, but `rollback --list` showed opaque IDs only and `Restored snapshot` did
+      not disclose that zero effective files changed. Add timestamp, mutation kind,
+      affected scope, and no-op/current-state guidance; add a regression journey with
+      two adjacent snapshots and explicit older-snapshot rollback.
+- [ ] `P18-22 [TERRA]` Remove the dashboard before the public release. Founder review
+      confirmed that it presents recommendation presets (`stable`, `web`,
+      `collaboration`, `maximum`) as policy profiles while the real CLI contract is
+      `stable`, `power`, `maximum`, and `custom`; its manifest-sync mutation model is
+      not the CLI setup workflow. Remove the command, loopback server, browser assets,
+      dashboard-only dependencies/tests/docs/evidence, and npm package contents.
+      Preserve any unique useful inspection capability through existing CLI commands.
+- [ ] `P18-23 [HUMAN+TERRA]` Complete the remaining CLI-only founder path on the exact
+      npm package. Power and its explicit rollback are complete: snapshot
+      `1784546929191-9cfb0705dbed` restored zero Loadout-managed activations, retained
+      all 12 unmanaged Claude skills, returned Codex to zero skills, and left no
+      duplicate groups. Maximum setup is also complete: snapshot
+      `1784547473319-7581eb19ee9c` stored 2,316 screened skill copies from 29 packages
+      in the disabled library, activated none of them, and again preserved the 12
+      unmanaged Claude skills. Remaining path: project activation -> recommendation/
+      optimization -> Graphify install/remove -> credential-free MCP inventory ->
+      read-only update/discovery -> complete uninstall -> reinstall. Record every
+      mutation snapshot ID and use explicit rollback IDs during acceptance.
+- [ ] `P18-24 [SOL+TERRA]` Redesign Power before 0.4.1 based on the real founder run.
+      The published transaction itself passed: eight immutable sources installed for
+      Claude Code and Codex, 100 managed activations were created, 12 unmanaged Claude
+      skills were preserved, duplicate targets were resolved, six rejected skill
+      units were excluded, and 1,170 managed files reported zero drift. The profile
+      policy failed acceptance: it activates 50 skills per agent despite Loadout's
+      recommended limit of 30; every selected package retained blocking static-risk
+      findings (79 total); only five of eight sources have an asserted SPDX license;
+      and one coarse prompt approves all package findings. Make Power a broad managed
+      library with no more than 30 project-relevant skills active per agent, preserve
+      per-package approval and rejection choices, persist and display excluded-unit
+      quarantine evidence, and make the health headline reflect blocking risk and
+      capacity failures instead of reporting the overall installation as healthy.
+      Add a release test proving Power preserves unmanaged skills, never exceeds its
+      active cap, reports every excluded unit and retained finding, and restores its
+      explicit setup snapshot without residue. The founder's published-package run
+      has now demonstrated that final rollback property successfully.
+- [ ] `P18-25 [LUNA+TERRA]` Make health output honest and focused for empty or unmanaged
+      profiles. After the successful Power rollback, `loadout health --explain` led
+      with `Loadout health: healthy` and then reported 28/100 critical evidence for
+      Claude Code, 0/100 unknown for Codex, and similarly verbose sections for every
+      detected agent. Use an explicit `not configured`/`no managed packages` overall
+      state instead of healthy, keep the default summary concise, and let users scope
+      detailed evidence to selected agents while preserving the full explanation on
+      request. Add empty-state and mixed detected-agent CLI regression tests.
+- [ ] `P18-26 [LUNA+TERRA]` Make Maximum's preview understandable without weakening
+      its safe defaults. The founder run proved the disabled-library contract, but
+      printed 50 unit-level quarantine blocks, 19 expected MCP/runtime deferrals, two
+      `No SKILL.md` preparation failures, and one coarse 28-package risk approval in
+      the default path. The subsequent scan reported 11 repository failures without
+      naming them. Default to a concise grouped summary with exact counts, severity,
+      source, and next commands; retain complete findings behind an explicit details
+      or JSON view. Classify no-skill MCP/runtime records as explicit setup rather
+      than failed skill preparation, distinguish high-confidence findings from likely
+      documentation/binary/metadata false positives, persist every excluded unit, and
+      name every failed repository with a short reason. Add snapshot tests for both
+      concise and detailed Maximum output.
+- [x] `P18-27 [SOL+TERRA, LUNA copy review]` Correct project-aware activation safety
+      and relevance using
+      `docs/superpowers/specs/2026-07-20-project-activation-safety-design.md`.
+      Founder preview proved that activation currently ignores 12 unmanaged Claude
+      skills when calculating `--limit 30`, treats rollback-restored empty directories
+      as occupied for both agents, and proposes a generic, redundant 30-skill set from
+      only JavaScript/TypeScript and Playwright signals. Implement the approved shared
+      empty-target predicate, per-agent total active capacity, bounded Node CLI/npm/
+      Vitest/MCP project signals, diverse evidence-threshold selection, integration-
+      type labels, atomic apply revalidation, and the specified regression suite.
+      Implementation is complete on `codex/project-activation-safety`: the full local
+      release gate passes, including 114 unit-test files with 605 passing tests and
+      one intentional skip; the disposable two-agent CLI journey proves 12 unmanaged
+      Claude skills, distinct Claude/Codex budgets, empty rollback residue, one atomic
+      apply, and explicit rollback; the packed artifact is 468.0 kB with 145 files;
+      and the fresh scan benchmark passed at 1,583.5 ms p95 across seven real CLI
+      runs. Do
+      not resume real-profile activation until this branch is reviewed, integrated,
+      versioned, and published as the corrected release.
 
 ### Release 0.3 lifecycle hardening
 
@@ -52,9 +210,9 @@ can test Loadout.
 - [x] `P18-12 [TERRA]` Add a pinned Chrome DevTools MCP recipe and distinguish
       separately billed AI/model API keys from unrelated service credentials. The
       no-model-key inventory includes GitHub read-only with its token disclosed.
-- [ ] `P18-13 [HUMAN+TERRA]` Publish `0.3.1`, then run Stable -> rollback -> Power ->
-      rollback -> Maximum -> project optimization -> dashboard -> complete uninstall
-      against the exact npm package.
+- [x] `P18-14 [TERRA]` Treat recursively empty directories from legacy cleanup as
+      unoccupied without weakening unmanaged-file protection, and make complete
+      uninstall remove empty nested directory shells.
 
 ### Explicitly deferred (do not expand during this usability pass)
 
@@ -66,6 +224,12 @@ can test Loadout.
 - A universal quality score, social-network scraping, or support claims for an agent
   that has not passed a real adapter test.
 
+## Archived implementation history
+
+Sections 1–20 below preserve the original product design, allocation, completed work,
+and deferred exploration. They are historical context. Only the current-status section
+above defines active work.
+
 ## 1. Executive summary
 
 Loadout is a universal extension manager for AI coding agents. It detects the agents