@pilotspace/add 1.13.0 → 1.15.0

package/CHANGELOG.md

@@ -4,7 +4,195 @@ All notable changes to the ADD method (`@pilotspace/add` on npm,
 `pilotspace-add` on PyPI) are documented here. The format follows
 [Keep a Changelog](https://keepachangelog.com/); versions follow semver.
 
-## [Unreleased]
+## [1.15.0] — 2026-07-02
+
+Ten milestones — the largest bundle yet — round out ADD's self-knowledge
+(**context-search**, **seams**, **artifact-graph**, **traceability-ids**,
+**ground-trust**, **drift-guard**), extend it past static rules
+(**persona-teacher-bundle**, **persona-learning-loop**, **advisor-gated-autonomy**),
+and past Claude Code itself (**portable-roster**). All additive; nothing removed
+or renamed on the CLI surface.
+
+### Added (context-search — find prior work before you drift into it)
+- **`add.py search <keyword...>`** — case-insensitive substring search over the
+  milestone/task corpus (active + archived), title/goal/rationale lines only, never
+  the full body. Surfaced at new-scope drafting and inside the specify/scenarios
+  phase guides, so related prior work is found before drafting — not after a
+  conflicting design ships.
+
+### Added (seams — one home for a shared contract)
+- **`SEAMS.md`.** Symbols that ≥2 tasks touch get promoted into a milestone-level
+  doc that §0 references, so a shared contract has one home instead of being
+  re-derived — and drifting — per task.
+
+### Added (artifact-graph — a traversable cross-artifact graph)
+- **Bidirectional backlinks.** Every ADD artifact now carries minimal backlink
+  metadata (task↔milestone↔release↔deps↔delta) so the cross-artifact graph is
+  traversable without re-deriving it by hand.
+
+### Added (traceability-ids — every rule earns a stable ID)
+- **`M#` / `R#` rule IDs.** Every §1 Must/Reject rule gets a stable ID that §2
+  scenarios and §4 tests reference via a `covers:` line, plus lint coverage so no
+  rule ships unscenarioed or untested.
+
+### Added (persona-teacher-bundle — a vendored, pinned teacher corpus)
+- **agency-agents, vendored.** The agency-agents teacher corpus ships as a pinned,
+  MIT-attributed local library (`.add/personas-teacher/`), bundled into both the
+  npm and PyPI distributions, de-branded from method prose, and refreshed on a
+  schedule. The engine stays NO-EXEC; the release build is zero-network.
+
+### Added (persona-learning-loop — personas that learn your project)
+- **Project-fit personas.** The AI seeds requirements personas at setup (a living
+  doc the project uses live), grows them through the observe→fold self-improve
+  loop, applies them at UDD/advisor/build, and exposes a cross-runner
+  (Claude Code · Codex · …) persona-aware subagent prompt.
+
+### Added (advisor-gated-autonomy — earn autonomy through instrumentation, not by removing the gate)
+- **A persisted, advisor-guarded `auto + parallel` run mode** so high-speed builds stay safe
+  without a human on every step. The whole milestone was dogfooded — every task built
+  `conservative` / human-gated, because you cannot auto-gate the feature that relaxes auto-gating.
+- **`sensitivity:` taxonomy.** A human-declared TASK-header risk-CLASS (base
+  `security · data · architecture · mechanical`, project-extensible via `GLOSSARY.md`) — ADD now
+  reasons about *what kind* of risk, not just *how much* (`risk:`). Validated at freeze; the engine
+  never classifies.
+- **Persisted run mode + setup choice.** The `auto + parallel` posture is persisted and chosen at
+  setup; `add.py status` shows `run mode: parallel + auto`. The non-interactive default stays
+  byte-identical.
+- **Persisted DAG-plan snapshot.** An auditable snapshot of the computed plan (waves · critical path ·
+  tiers) with a freshness check vs the live `depends_on` edges (edges stay the source of truth).
+- **Advisor 3-lens verdict.** The refute-read becomes a tier-aware single advisor running a sequential
+  security → concurrency → architecture sweep recorded in §6 (Verdict · Residue · Binding) — the
+  non-functional review is now structured and auditable; `advisor_verdict_unrecorded` measure-not-block lint.
+- **Advisor coverage audits (measure, never block).** `add.py audit` flags a missing advisor verdict,
+  a self-reviewed verdict (reviewer = author), and a mechanical task with advisor-found residue (mis-tier).
+- **`advisor-gate-relax` — the narrowest safe relaxation.** A `risk: high` + `sensitivity: mechanical`
+  task with a recorded `Verdict: PASS` + `Residue: none` auto-completes via `gate PASS` without a lowered
+  autonomy level. **Security and every non-mechanical class are never relaxed** — the human-in-the-loop
+  floor is untouched.
+- **Per-phase spawn hint.** `status`/`guide` print an advisory subagent-spawn hint (idiom + tier) for the
+  active phase; advisory only — the engine still never spawns.
+- Documentation, glossary (4 new terms), and headers aligned with the feature.
+
+### Added (portable-roster — the phase-roster for every coding tool, not just Claude)
+- **`AGENTS.md` / `.clinerules` carry the roster.** Non-Claude coding tools (Cursor,
+  Copilot, Codex, Cline, …) now receive the ADD phase-roster's 5 roles and
+  boundaries through the `AGENTS.md` the installer already drops, not just
+  Claude Code's native subagents.
+- **`add-advisor` — the 5th roster agent.** A consultative, frontier-model agent
+  (`model: opus`) any phase can spawn on a medium-hard decision — an ambiguous
+  read, a risky shape, a change of approach. It recommends and weighs tradeoffs;
+  it never decides.
+
+### Changed (ground-trust — GROUND surfaces problems, not just structure)
+- **GROUND now surfaces the issues/risks** it finds in the real code (feeding
+  SPECIFY) and links each task's related intent back to the foundation
+  (`PROJECT.md` · `GLOSSARY.md` · conversation) — specs build on problems found,
+  not assumed.
+
+### Changed (drift-guard — kill §0 reference rot)
+- **§0 cites symbols, not line numbers**, stamps `ground_sha`, refreshes at close,
+  and strips dead live-phase scaffolding at done — a closed TASK.md stays true to
+  the code instead of rotting the moment a line shifts.
+
+### Changed (loop-readability — human-scannable output across every phase)
+- **Report shape, audited and tightened.** `report-template.md`'s ARC + 6-block
+  shape was checked against its own stated rules (summary-first, one decision,
+  guided-choice) across all 8 phase guides' gate-reporting cues; 2 guides
+  (`0-setup.md`'s baseline-lock sentence, `3-contract.md`'s freeze-gate
+  sentence) had drifted and were tightened to name SHAPE ahead of the APPROVE
+  guided choice.
+- **SKILL.md's compact pipeline sentence now names the decision banner**
+  ("rendered first, above everything") ahead of the ARC, matching
+  `report-template.md`'s actual render order — closing a gap the shape audit
+  surfaced.
+- **`MILESTONE.md.tmpl` gains a UI/UX Scope hint** — the axes to name
+  precisely for a UI feature (information architecture, interaction pattern,
+  visual hierarchy, design tokens, component states, accessibility floor,
+  responsive breakpoints, user journey), pointing at `design.md` — with
+  adoption pointers wired into `1-specify.md` and `scope.md` so drafters
+  actually reach it. The seed this milestone was created to dogfood.
+
+### Added / Changed (loose tasks since 1.14.0)
+- **Report template PLAN/SHAPE + APPROVE banner** (`report-plan-approve`) — the
+  chat-report decision point renders a clearer guided-choice banner.
+- **`status` pagination** (`status-pagination`) — milestones/tasks sort by
+  updated-descending and cap to the top 10, with a `--all` escape hatch, so a
+  long-lived project's `status` stays scannable.
+- **Lean 3-agent phase roster + adaptive persona agent** (`phase-agents-lean`).
+- **Skill-tree compaction** (`skill-tree-compaction-audit`) — audited and
+  compacted the ADD skill tree for genuine prose redundancy under the pinned
+  lean-fence budget.
+- **Vendor-tree `.gitignore` fixes** — nested `.add/.gitignore` vendor-tree
+  patterns now resolve relative to `.add/`, not repo root
+  (`gitignore-vendor-path-fix`); the installer's own `.gitignore` seed covers all
+  3 managed vendor trees (`installer-gitignore-mirrors`); `update --global`
+  re-seeds `.gitignore` for every registered project, not just fresh installs
+  (`update-global-gitignore-seed`).
+- **CI tooling-mirror gap closed** (`ci-tooling-mirror-gap`) — the `.add/tooling`
+  dogfood mirror is materialized in CI's test + publish-guard jobs, not just
+  locally.
+- **Fresh-checkout skip-count tolerance widened**
+  (`nested-suite-skip-count-tolerance`) — the nested-suite OK-regex now tolerates
+  every known environment-conditional skip, not just the recursion guard.
+- **Scope-walk hygiene** — `.claude` pruned from the scope walk
+  (`scope-exclude-claude`); stale mirror trees re-synced to canonical
+  (`mirror-resync`); `.add/tooling` untracked as a regenerable dogfood mirror
+  (`untrack-add-tooling`).
+
+### Changed
+- Five version sources bump in lockstep to **1.15.0** (`package.json`,
+  `package-lock.json` ×2, `pyproject.toml`, `.claude-plugin/plugin.json`,
+  `add_method.__version__`).
+
+This release bundles **11 closed milestones** (`seams`, `context-search`,
+`drift-guard`, `artifact-graph`, `ground-trust`, `traceability-ids`,
+`persona-teacher-bundle`, `persona-learning-loop`, `advisor-gated-autonomy`,
+`portable-roster`, `loop-readability`) and 13 loose tasks since 1.14.0. Every
+milestone was built end-to-end through ADD's own spec→tests→build→verify flow.
+25 open SPEC deltas (non-security backlog spanning the newly-grounded
+install-update-hardening tasks, report-shape-scan-audit/skill-banner-cue
+follow-ups, and pre-existing items — full list: `add.py deltas`) ride forward
+unresolved into the next cycle.
+
+## [1.14.0] — 2026-06-29
+
+Two milestones round out lanes that shipped partially in earlier releases: the
+**global-home / installer** lane gets its missing inverse + safety, and the
+**component** pillar closes its remaining gaps and hardens its cross-repo edges.
+Installer- and engine-pin-neutral — the ADD engine is byte-identical (ENGINE_MD5
+unchanged); all changes live in the installer twins and the component validator.
+Backward-compatible throughout.
+
+### Added (installer-polish — complete the global lane)
+- **Restore the global home.** `init --from-global-data` (and an `init` that detects a
+  matching `<home>/data/<key>`) rehydrates a project's user-data from the global home on
+  a fresh clone — the non-destructive inverse of the one-way backup. Fill-gaps by default;
+  `--force` writes a `<name>.bak` sidecar before overwriting.
+- **`prune-data` orphan cleanup.** A new `prune-data` command removes home snapshots with
+  no live registry owner — dry-run by default, `--force` to delete. Both installer twins
+  (pip + npm) carry the behavior, byte-for-byte.
+- **`update --global` made safe.** An O_EXCL home lock (`<home>/.update.lock`) serializes
+  concurrent runs **cross-twin** (a pip-held lock blocks an npm run and vice-versa), and
+  every registered path is validated before any write — a relative/traversal entry aborts
+  the whole run loud (`unsafe_registry_path`), a directory without `.add/` is dropped.
+- **Reconcile roll-up.** Every reconcile now reports a file-level `N restored · M refreshed`
+  summary, so a partially-gutted-but-present managed tree's heal is finally visible (it
+  healed silently before). Pure observation — copy semantics are unchanged.
+
+### Added (component-polish — close the pillar gaps + harden the edges)
+- **`add.py components` reader + validator.** A `components.toml` schema-lint surfaces
+  `component_unknown_key` / `component_type_mismatch` / `component_unknown_table` — all
+  measure-not-block warnings at `check`.
+- **Federation hardening.** `federate pull` path-confines the manifest `source` to a
+  sibling-repo allowlist with a fail-closed HARD-STOP (`federation_source_escapes`) before
+  any read; a stale leftover contract snapshot that no longer admits a consumer is surfaced
+  (`producer_contract_stale`, never red).
+- **Registry fill + a worked example.** The component registry round-trips completely, and
+  a full worked example threads the component flow end-to-end in the book.
+
+### Changed
+- Five version sources bump in lockstep to **1.14.0** (`package.json`, `package-lock.json`
+  ×2, `pyproject.toml`, `.claude-plugin/plugin.json`, `add_method.__version__`).
 
 ## [1.13.0] — 2026-06-28
 

package/THIRD_PARTY_NOTICES.md

@@ -0,0 +1,41 @@
+# Third-party notices
+
+This project bundles third-party material. Each component below is redistributed under its own
+license, retained in full alongside the vendored files. These notices satisfy the attribution
+requirement of the licenses; they are not affected by any product/marketing naming choices.
+
+---
+
+## Vendored teacher corpus — `add-method/personas-teacher/`
+
+A pinned snapshot of an external library of agent definitions, vendored as a local teacher corpus
+that the persona phase reads off-build. The pinned upstream commit and trim rules are recorded in
+`add-method/personas-teacher/VENDOR.md`; the upstream license is retained verbatim at
+`add-method/personas-teacher/LICENSE`.
+
+- License: **MIT**
+- Copyright (c) 2025 AgentLand Contributors
+
+```
+MIT License
+
+Copyright (c) 2025 AgentLand Contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+```