@antoneeo/agentic-sdlc-skill 1.11.0 → 1.12.0

package/CHANGELOG.md

@@ -1,227 +1,239 @@
-# Changelog - Agentic SDLC Skill
-
-Tutte le modifiche significative a questa skill saranno documentate in questo file.
-
-## [1.11.0] - 2026-07-03 (M4: Consolidation & Proactive Activation — self-activating, self-consulting, self-testing skill)
-### Added
-- **SessionStart orientation hook** (`sdlc_check.py orient`): emits a bounded, repo-sourced orientation (README + INDEX + guide router + handoff + Rule-Zero triage) at session start. Zero-execution, **fail-OPEN** (a missing/empty `ai_docs/` never blocks the session), size-capped. Manual per-client wiring in `ENFORCEMENT.md` §4; `--hybrid` points at the devPNT bootstrap instead of duplicating plan/KL. `test_session_start.py` (9 cases).
-- **Guide-layer consumption** (closes the write-only gap — Layer D "point to them"): a **consult trigger** (before operative L2/L3 work, targeted router match, L1 exempt, never blanket) and a **proactive-creation trigger** (propose a guide after user-indication-governed reusable work; never silent, never from model knowledge). Mechanics in `guides.md` §0/§1; hooked from `SKILL.md` Operative Guides + Phase 4/5; reconciled with subagent dispatch in `dispatch.md`.
-- **Worktree/branch hygiene** in the closure discipline (`SKILL.md` Phase 4 isolate-on-branch / Phase 5 merge-decision + cleanup).
-- **Skill eval harness** (dev-only, not shipped): `test_skill_invariants.py` is the deterministic static release gate (`python -m unittest discover -s scripts -p "test_*.py"` — asserts the skill's own doctrine invariants: triggers/hook/worktree present and wired, indexes idempotent, support pointers resolve; zero LLM/network/subprocess). Opt-in behavioral corpus `evals/scenarios/` + `run_behavioral.py` (non-CI, never gates). `ENFORCEMENT.md` §5.
-
-### Process note
-- Full governance per unit: M-VISION v2.1 (revised — added the guide-consumption unit) → D-UC + P-TM → per-unit E-ISP (deep review) + E-TDD (light review) → implement → §4.6 code review. The independent-review gate caught real defects at design time (incomplete eval invariant set, the M3↔M4 dispatch interaction, a REPO path off-by-one, a P-TM overclaim of unbuilt guards). Battery 51/51 green. ADR `adr_2026-07-03_skill_eval_harness`; KL architecture v1.4 + principles v1.2.
-
-## [1.10.0] - 2026-07-03 (M3: Subagent Execution / Feature A — opt-in executable plan)
-### Added
-- **`dispatch.md`**: subagent-execution doctrine (opt-in for L3). The dispatch loop — validate the plan → per-task brief → economy-tier implementer → one-shot review → ledger — with client-relative model tiers (no provider names), one-shot review slots (no iterative loops), degradation to same-session where subagents do not exist, and guides injected by pointer (never pasted).
-- **`sdlc_check.py plan` subcommand**: `plan validate` (schema check of the executable plan, fail-closed path/guide confinement, sidecar-ledger cross-check — "no valid plan, no dispatch") and `plan brief --task <id>` (emits, to stdout, the task + prior-task interfaces + guide pointers). The validator is **zero-execution**: a task's `verify` command is emitted as text, never run.
-- **Executable-plan template** (`ai_docs/solutions/PLAN_[feature].md`) in `templates.md`: Markdown frontmatter (`status`, `derived-from`) + a fenced `json` task array + the sidecar `PLAN_[feature].ledger.json` shape (`task_id -> {status, verify_result, timestamp}`, git-tracked, survives compaction).
-- **`SKILL.md` §4** opt-in subagent-execution hook + the Hybrid `derived-from` seam (the plan is derived from the accepted E-TDD, never independently authored).
-- `test_plan.py`: stdlib-`unittest` battery for the `plan` subcommand (32 cases: schema, confinement, ledger, fail-fast JSON, zero-execution poka-yoke).
-
-### Changed
-- **`confine_under(base, rel)` extracted** in `sdlc_check.py`: the fail-closed path-confinement pattern (absolute/`..`/resolve-escape → reject), previously inlined twice (the `overrides:` and `distilled_from` checks), is now a single helper reused by both plus the new plan-path / guide-pointer confinement. Behavior-preserving (catches `(ValueError, OSError)`).
-
-### Process note
-- 4th live **model-per-dispatch** run (economy implementer from the E-TDD shadow, battery 32/32, deep code review PASS zero BLOCK). Governance: M-VISION → D-UC → P-TM → E-ISP → E-TDD, all through the independent review gate — which killed 3 real BLOCKs at design time (T1 subprocess-invariant misstatement, a missing impacted file, a `confine_under` OSError-crash regression). ADR `adr_2026-07-03_executable_plan_json_in_md`.
-
-## [1.9.0] - 2026-07-03 (M2 execution disciplines + Feature B unit 2 agent KB)
-### Added
-- `tdd.md`: TDD discipline (RED/GREEN/REFACTOR, increment rule, AAA test shape, documented exemptions) — the L2/L3 default for implementation work.
-- `debugging.md`: systematic debugging method (reproduce, isolate, root cause, fix, regression test, collateral check) with circuit-breaker integration.
-- `elicitation.md`: spec elicitation round (goal/benefit, scope boundaries, non-goals, constraints, acceptance signals) run before drafting the analysis in phase 3.
-- `review.md`: single definition of requesting, receiving, and performing code review, reused by the Hybrid review gates.
-- `SKILL.md` wiring: pointers to the four new files added at phase 3 (elicitation), phase 4 (TDD, debugging, circuit breaker), and phase 5 (review).
-- Agent-global KB (fixed root, project-wins precedence, `overrides:` with fail-closed confinement, collision warnings).
-
-## [1.8.1] - 2026-07-02
-### Fixed
-- **Guide freshness hash is now line-ending independent**: `sha256_file` in `sdlc_check.py` normalizes CRLF → LF before hashing. Previously the raw-byte hash made a fresh Windows checkout with `core.autocrlf=true` rewrite `.sources/` snapshots and flag every guide `[stale]` (false positive). Backward compatible: recorded hashes were computed on LF content, and normalization maps CRLF copies back to the same digest. (Edge case: a hash recorded pre-1.8.1 on a snapshot that genuinely contained CRLF bytes will flag `[stale]` once — regenerate the hash.)
-- `guides.md` step 3 now states the hash is computed over LF-normalized content and recommends the `ai_docs/reference/.sources/** -text` `.gitattributes` rule to consumer projects (defense in depth: keeps snapshots byte-verbatim).
-
-### Test battery addition
-- Scenario 10 (extends the unit-1 battery): snapshot checked out with CRLF endings + guide recording the LF-normalized hash → `stale` must NOT flag it; a genuine content edit must still flag `[stale]`.
-
-## [1.8.0] - 2026-07-02 (Feature B unit 1: operative guides, project scope)
-### Added
-- **Operative guides** (`ai_docs/reference/GUIDE_[topic].md`): a durable, source-faithful layer distilled from USER-PROVIDED indications — the capability neither agentic-sdlc nor superpowers had. New support file `guides.md` (pipeline: topic decomposition → user confirmation → verbatim snapshot in `reference/.sources/` with SHA-256 → source-anchored extraction → per-section fidelity markers `[source: …]` / `[not covered by source]`); guide template with provenance frontmatter (`source`, `source_version`, `distilled_from`, `source_hash`) in `templates.md`; short "Operative Guides" section in `SKILL.md`.
-- **Generated guide router** `ai_docs/reference/INDEX.md` (emitted by `sdlc_check.py index`, alignment-checked by `validate`): path, status, when-to-consult line and provenance summary per guide — the pointer target for the devPNT Hybrid bridge.
-- **Mechanical fidelity controls** in `sdlc_check.py`: provenance-key and per-section marker checks (warn; fails CI under `--strict`); `distilled_from` path confinement — absolute paths, `..` and symlink escapes rejected, fail-closed (ERROR); guide freshness in `stale` — recorded `source_hash` vs current snapshot, flagged in EVERY mode including `--hybrid` (guides are filesystem-first even in Hybrid).
-
-### Changed
-- `stale --hybrid` no longer returns unconditionally 0: it still skips audit-plan staleness (delegated to devPNT/KL) but now checks guide-source drift.
-- `list_canonical_docs` skips dot-subdirectories of canonical dirs (e.g. `reference/.sources/`): snapshot files are no longer swept into the manifest. Projects that kept `.md` files under dot-subdirs of canonical dirs will see them leave `INDEX.md` (more correct).
-
-### Process note
-- First live run of **model-per-dispatch**: implementation dispatched to an economy-tier subagent working from the accepted E-TDD shadow as a self-contained brief; independent deep code review passed first round with zero blocking findings. Governance: M-VISION → D-UC → P-TM → E-ISP → E-TDD, all through independent review gates (devPNT Hybrid).
-
-## [1.7.0] - 2026-07-02 (Phases 0-1 of the evolution roadmap)
-### Changed (breaking-soft)
-- **English is now the canonical language of the skill**: `SKILL.md`, `templates.md`, `ENFORCEMENT.md`, validator messages, generated indexes and the project protocol are in English. New ANALYSIS documents use English frontmatter keys (`status`, `level`, `start_date`, `end_date`) and English section headings. **Existing projects keep working**: the validator silently accepts the deprecated Italian keys (`stato`, `livello`, `data_inizio`, `data_fine`) and Italian headings.
-- The generated project protocol (`CLAUDE.md`/`GEMINI.md`/`AGENTS.md`/`.cursorrules`) is now a **thin pointer** to the skill (triage summary + where things live + closure gate) instead of a condensed copy of its rules, which had drifted from `SKILL.md`.
-- `init.js` now seeds `ai_docs/` from `templates.md` (single template source) instead of inline boilerplates; it creates `ai_docs/reference/` and the curated `ai_docs/README.md`, no longer seeds the generated `features_history.md`, and generates `INDEX.md` via the validator when Python is available, so the very first `sdlc_check.py check` on a fresh project is CLEAN.
-- Client detection unified between `init.js` and `postinstall.js` (`scripts/lib.js`): CLI on PATH **or** config home present (covers Claude Desktop with integrated Claude Code).
-
-### Added
-- `sdlc_check.py validate --strict` / `check --strict`: warnings and a missing `ai_docs/` become failures (for CI).
-- **Coexistence with devPNT (Phase 1, the Hybrid seam)**: new SKILL.md section with the ownership matrix (who is master per artifact in Standalone vs Hybrid), the triage equivalence table (one significance threshold, two vocabularies), the ANALYSIS↔plan-node state mapping and the shadow discipline (`SHADOW_[doc_key]_vX.Y.md`, exported BEFORE implementation; never saved under an `ANALYSIS_*` name).
-- `sdlc_check.py --hybrid` (explicit, never auto-detected) on `check`/`stale` (audit-plan staleness delegated to devPNT/KL) and on `gate` (an approved E-TDD shadow in `solutions/` authorizes writes on protected paths).
-- devPNT MCP doctrine (`mcp_system_prompt.md`, devPNT repo) slimmed accordingly: process (triage, phases, lifecycle, closure) deferred to the skill; `ai_docs/` layout aligned (adds `vision/` and `reference/`); shadow naming and shadow-before-implementation rule; checklist items for methodology, shadow export and hybrid closure gate.
-- Evolution roadmap for v1.7.0 in `ai_docs/vision/roadmap_evoluzione_agenti.md` (subagent execution, operative guides + agent-level knowledge base, devPNT seam, open-core positioning).
-
-### Fixed
-- Shadow detection is structural (filename `SHADOW_*` or `<!-- SHADOW` marker on the first line): an ANALYSIS merely *mentioning* shadows is no longer silently skipped from index and validation.
-- `mark`/`index` fail fast when `ai_docs/` is missing instead of silently creating a second documentation root in the wrong directory.
-
-### Removed
-- Divergent `agentic-sdlc-v2/` copy (integrated in 1.5.0, the leftover risked edits on the wrong files).
-- `references/*_template.md` (duplicated `templates.md` and had already diverged).
-
-## [1.6.0] - 2026-06-15
-### Added
-- **Manifest generato dei documenti canonici** (`ai_docs/INDEX.md`): `sdlc_check.py index` ora produce, oltre a `features_history.md`, un indice completo di tutti i doc in `vision/`, `reference/`, `architecture/`, `functional/`, `strategic/`, con descrizione e stato letti dall'header. Si rigenera, quindi non drifta.
-- **Lifecycle dei documenti canonici**: convenzione header `status: CURRENT|SUPERSEDED|DRAFT|DEPRECATED` + `supersedes:`. `validate` avvisa se `status` manca/è invalido o se un doc superseduto è ancora `CURRENT`. Stop ai grep che riportano a guide obsolete.
-- **Modello a due indici** documentato nella sezione "Documenti ai_docs" di `SKILL.md`: `README.md` curato (must-read, a mano) vs `INDEX.md` generato (completo, meccanico) — ruoli separati, prima confusi in un unico README che driftava.
-
-### Changed
-- §1 Audit: leggere `README.md` + `INDEX.md` all'avvio per sapere cosa esiste prima di esplorare il codice.
-- §3 Analisi: cercare con glob/grep un'ANALYSIS esistente prima di crearne una nuova (anti-duplicazione).
-- §5 Chiusura: gate "Indici allineati" — rigenerare `INDEX.md`, aggiornare il `README.md` curato per i must-read, marcare lo `status`; doc canonico non indicizzato o senza `status` = chiusura sporca.
-- `templates.md`: aggiunto il template dell'header dei documenti canonici.
-
-### Fixed
-- `sdlc_check.py` legge ora i file con `utf-8-sig`: un BOM iniziale (file autorati su Windows) non impedisce più il riconoscimento del frontmatter `---`.
-- L'estrattore dell'header riconosce sia il frontmatter `status:` sia la riga in corpo `**Status:**`/`Stato:`, e gli stati di tutte le convenzioni in uso (canonici `CURRENT/SUPERSEDED/DRAFT/DEPRECATED`, vision `DRAFT/APPROVED`, ADR `Accepted/Proposed/Rejected`) — niente più falsi avvisi "status non riconosciuto" su `APPROVED`/`Accepted`.
-- La descrizione del manifest salta righe di metadati (`Date`, `Created`, `Task ref`, ...) e i commenti HTML, così non finiscono come descrizione del documento.
-- `index` non genera più un `INDEX.md` vuoto su progetti senza documenti canonici (solo `solutions/`+`audit/`).
-
-### Migrazione (da 1.5.x)
-- Al primo `sdlc_check.py check`/`validate` dopo l'upgrade, un progetto con documenti canonici darà **un errore** `ai_docs/INDEX.md mancante`: è atteso — esegui **una volta** `sdlc_check.py index` per generarlo. Da lì in poi resta allineato.
-- I documenti canonici preesistenti senza `status:` produrranno **avvisi** (non errori): aggiungi l'header `description:`/`status:` per silenziarli. I nuovi progetti nascono già compatibili (template aggiornati).
-
-## [1.5.0] - 2026-06-13
-### Added
-- Introdotta la Regola Zero di triage (`L1`, `L2`, `L3`, `Spike`) per rendere il processo proporzionale al rischio.
-- Aggiunta simbiosi esplicita con devPNT: in Hybrid la `M-VISION` guida la milestone, il Master Plan resta roadmap strategica e l'Action Plan governa l'esecuzione tattica.
-- Aggiunti support file dentro la skill runtime: `templates.md`, `ENFORCEMENT.md`, `scripts/sdlc_check.py`.
-- `agentic-sdlc-install-skill` ora installa la skill nativa anche in `~/.gemini/skills/agentic-sdlc/`.
-- Aggiunto validatore meccanico opzionale per frontmatter ANALYSIS, Vision state, indice feature e audit stale.
-
-### Changed
-- Il nome pubblico resta `agentic-sdlc`; la proposta v2 e' stata integrata come evoluzione, non come skill parallela.
-- Aggiornati `agentic-sdlc-init`, template, protocolli generati, README e metadata.
-- La modalita Standalone resta completa; devPNT e' un livello di governance superiore, non un prerequisito.
-
-## [1.4.0] - 2026-06-07
-### Added
-- Introdotta la governance della **Vision** con nuova struttura `ai_docs/vision/` (`project_vision.md`, `roadmap.md`, `principles.md`, `features/`).
-- Aggiunto il **Vision Gate** nel workflow operativo: ogni feature significativa deve essere verificata rispetto a obiettivi, non-obiettivi, benefici attesi e segnali di successo prima dell'analisi tecnica.
-- Aggiunti template Vision in `references/` e sezione `Allineamento alla Vision` nel template di analisi.
-- `agentic-sdlc-init` ora crea i documenti Vision boilerplate nei nuovi progetti.
-
-## [1.3.1] - 2026-05-14
-### Fixed
-- Correzione documentazione (README + CHANGELOG) della sintassi per invocare il bin `agentic-sdlc-install-skill`. La forma `npx @antoneeo/agentic-sdlc-skill agentic-sdlc-install-skill` documentata in 1.3.0 **non funziona** perché npx non riesce a disambiguare il bin quando il pacchetto ne espone più di uno (errore: `could not determine executable to run`). Sintassi corretta: lanciare `agentic-sdlc-install-skill` direttamente dopo `npm install -g`, oppure usare `npx -p @antoneeo/agentic-sdlc-skill agentic-sdlc-install-skill` con `-p` esplicito.
-- Nessuna modifica al codice della skill: il bin di 1.3.0 funziona correttamente, era solo la doc a indicare la sintassi sbagliata.
-
-## [1.3.0] - 2026-05-14
-### Added
-- Nuovo comando esplicito `agentic-sdlc-install-skill` (registrato come `bin`): installa la skill in `~/.claude/skills/agentic-sdlc/` e `~/.codex/skills/agentic-sdlc/` **senza dipendere dal `postinstall` hook**. Risolve i casi in cui `npm` salta gli script (configurazioni `ignore-scripts=true`, policy IT aziendali, alcuni installer Node) e l'auto-install fallisce silenziosamente.
-- Sezione "Troubleshooting" nel README con istruzioni per il caso in cui la skill non venga rilevata da Claude Code dopo `npm install -g`.
-
-### Usage
-```bash
-npm install -g @antoneeo/agentic-sdlc-skill@latest
-agentic-sdlc-install-skill
-```
-
-## [1.2.4] - 2026-05-14
-### Fixed
-- `postinstall` ora rileva Claude Code anche quando il CLI `claude` non è nel PATH (es. Claude Desktop con Claude Code integrato): la presenza di `~/.claude/` o della variabile `CLAUDE_CONFIG_DIR` è sufficiente per attivare l'installazione della skill in `~/.claude/skills/agentic-sdlc/`. Stesso pattern già usato per Codex.
-- `preuninstall` rispetta `CLAUDE_CONFIG_DIR` per rimuovere la skill dalla directory di configurazione corretta.
-
-## [1.2.2] - 2026-05-10
-### Changed
-- README convertito in inglese per la pubblicazione npm.
-- Versione allineata a `1.2.2` in `package.json` e `gemini-extension.json`.
-
-## [1.2.1] - 2026-05-10
-### Fixed
-- `postinstall` ora registra la skill anche per Codex copiandola in `$CODEX_HOME/skills/agentic-sdlc/` oppure `~/.codex/skills/agentic-sdlc/`.
-- `preuninstall` rimuove anche la copia Codex della skill.
-- `agentic-sdlc-init` ora crea `AGENTS.md` per Codex invece di `.codex/hooks.json`, che Codex non carica come istruzioni di progetto.
-
-## [1.2.0] - 2026-05-10
-### Added
-- **Auto-installazione skill nativa Claude Code**: il `postinstall` ora copia `skills/agentic-sdlc-skill/` in `~/.claude/skills/agentic-sdlc/` quando rileva il CLI `claude`. La skill diventa disponibile come `agentic-sdlc` nel tool `Skill` di Claude Code dopo riavvio.
-- Nuovo script `preuninstall.js` che rimuove la skill da `~/.claude/skills/agentic-sdlc/` durante `npm uninstall`.
-- Fallback `copyRecursive` per Node < 16.7 (quando `fs.cpSync` non disponibile).
-
-### Changed
-- `package.json`: bump versione a 1.2.0, aggiunti keyword `claude-code` e `claude-skill`, descrizione aggiornata per riflettere supporto Claude Code nativo.
-
-### Fixed
-- Risolto bug per cui il pacchetto npm non registrava la skill in Claude Code (la cartella `~/.claude/skills/` non veniva mai popolata).
-
-## [1.1.0] - 2026-05-10
-### Fixed
-- Fixed path inconsistencies in the generated `CLAUDE.md`, `GEMINI.md`, and `.cursorrules` protocols. Added full paths (`ai_docs/strategic/`) to all document references to ensure AI agents (like Claude) can correctly find and update them.
-- Cleaned up encoding issues in initialization scripts.
-
-## [1.0.9] - 2026-05-10
-### Added
-- New **Smart Discovery** system during project initialization.
-- Official support for **Codex AI** via automatic `.codex/hooks.json` injection.
-- Support for **Cursor** and **Windsurf** via `.cursorrules`.
-- Binary command `agentic-sdlc-init` for use via `npx`.
-- `postinstall` script for global AI CLI detection (Claude, Gemini, Codex).
-- Full English localization for all user-facing messages and scripts.
-
-### Changed
-- Updated `init.js` to support the new `ai_docs/` directory structure defined in v1.0.8.
-
-## [1.0.8] - 2026-05-10
-### Aggiunto
-- Skill evoluta in "Hybrid Edition": integrazione opzionale con **devPNT** (server MCP) per governance avanzata tramite database e piani gerarchici.
-- Fase di Discovery per il rilevamento automatico dell'ambiente (Standalone vs Hybrid).
-- Supporto per ADR (Architecture Decision Records) e Knowledge Layer (KL) nella fase di chiusura.
-
-### Modificato
-- Riorganizzazione dei percorsi di documentazione (`ai_docs/strategic/`, `ai_docs/audit/`, `ai_docs/solutions/`).
-- Aggiornata la documentazione funzionale (`architecture_overview.md`, `external_interfaces.md`) con definizioni più precise.
-
-## [1.0.7] - 2026-05-10
-### Modificato
-- Aggiornata l'attribuzione dell'autore (Antonio Pinto) e il copyright in tutti i file (`package.json`, `README.md`, `SKILL.md`).
-- Aggiunto link al profilo GitHub ufficiale.
-
-## [1.0.6] - 2026-05-10
-### Aggiunto
-- Sincronizzazione completa dei file del progetto nel repository.
-- Supporto per il tracciamento delle feature e audit plan.
-
-## [1.0.5] - 2026-04-17
-### Aggiunto
-- File `CHANGELOG.md` per il tracciamento delle versioni.
-
-## [1.0.4] - 2026-04-17
-### Modificato
-- `README.md`: Aggiunte istruzioni specifiche per l'installazione da chiavetta USB e chiarimenti sulla disponibilità globale della skill.
-
-## [1.0.3] - 2026-04-17
-### Aggiunto
-- Primo `README.md` con istruzioni di installazione e attivazione.
-
-## [1.0.2] - 2026-04-17
-### Corretto
-- Riorganizzata la struttura delle cartelle secondo gli standard di Gemini CLI (`skills/agentic-sdlc-skill/SKILL.md`).
-- Aggiunto frontmatter YAML a `SKILL.md` per la scoperta automatica.
-
-## [1.0.1] - 2026-04-17
-### Corretto
-- Aggiunto `gemini-extension.json` (manifest dell'estensione) mancante nella versione iniziale.
-
-## [1.0.0] - 2026-04-17
-### Iniziale
-- Prima pubblicazione della skill (Protocollo SDLC Documentation-First).
+# Changelog - Agentic SDLC Skill
+
+Tutte le modifiche significative a questa skill saranno documentate in questo file.
+
+## [1.12.0] - 2026-07-03 (Client Roster: Google Antigravity 2.0 as a first-class client)
+### Added
+- **Google Antigravity 2.0 support** in the install engine. The runtime skill now lands where all three Antigravity products (desktop, the `agy` CLI, the agentic IDE) discover global agent skills: `~/.gemini/config/skills/agentic-sdlc/`. `agentic-sdlc-init` writes the Antigravity project pointer to `AGENTS.md` (the Antigravity CLI surface), reusing the single `protocolContent` — no per-client drift.
+- **Shared-home collision resolved** (`~/.gemini`). Antigravity's global skills root lives UNDER `~/.gemini`, the home the legacy `gemini` client claims. A new distinct `antigravity` CLIENTS entry de-collides via two backward-compatible registry generalizations: an optional `skillsSubdir` (`config/skills`) on `skillTarget`, and an optional `homeMarker` (`~/.gemini/config/skills`) on `clientDetected` so Antigravity is detected only by its own skills dir, the `agy` CLI, or `ANTIGRAVITY_HOME` — never on bare `~/.gemini`. The existing three clients omit both fields and are byte-identical (no regression). See ADR `adr_2026-07-03_antigravity_gemini_home_collision`.
+- **Node test battery** `scripts/test_clients.js` (dev-only, NOT shipped — excluded from the `package.json` `files` allowlist, which is now an explicit per-file list of the four lifecycle scripts). 8 cases covering the P-TM threats: T1 shared-home double-install (bare `~/.gemini` → gemini TRUE / antigravity FALSE), T2 distinct skill-target + install/uninstall round-trip, T3 detection matrix (marker OR env OR CLI), T7 the three existing clients unchanged.
+
+### Changed
+- `package.json` `files` allowlist: the wholesale `"scripts"` directory entry is replaced by the explicit four shipped lifecycle scripts (`lib.js`, `init.js`, `postinstall.js`, `preuninstall.js`), so the dev-only `test_clients.js` is never packaged (the same dev-only precedent as the Python `test_*.py` batteries).
+
+### Process note
+- Full Hybrid governance (devPNT): M-VISION → D-UC → P-TM → E-ISP → E-TDD → ADR, all governed. Implemented against the accepted **E-TDD** `e_tdd_antigravity_client` v1.0 (shadow exported to `ai_docs/solutions/` before coding). TDD: detection/target logic tests-first (RED→GREEN); prose/doc edits exempt (recorded). `postinstall.js`/`preuninstall.js` unchanged (pure registry consumers). Node battery 8/8 + skill eval battery 51/51 green; `check --hybrid` CLEAN. Owner inputs resolved in round 2: CLI binary `agy`, env override `ANTIGRAVITY_HOME`, Antigravity-specific reload string. The repo doc `ai_docs/solutions/antigravity_skills_guide.md` (plugin/mcp_config model) is superseded by the accepted skills-model decision (E-ISP/ADR).
+
+## [1.11.0] - 2026-07-03 (M4: Consolidation & Proactive Activation — self-activating, self-consulting, self-testing skill)
+### Added
+- **SessionStart orientation hook** (`sdlc_check.py orient`): emits a bounded, repo-sourced orientation (README + INDEX + guide router + handoff + Rule-Zero triage) at session start. Zero-execution, **fail-OPEN** (a missing/empty `ai_docs/` never blocks the session), size-capped. Manual per-client wiring in `ENFORCEMENT.md` §4; `--hybrid` points at the devPNT bootstrap instead of duplicating plan/KL. `test_session_start.py` (9 cases).
+- **Guide-layer consumption** (closes the write-only gap — Layer D "point to them"): a **consult trigger** (before operative L2/L3 work, targeted router match, L1 exempt, never blanket) and a **proactive-creation trigger** (propose a guide after user-indication-governed reusable work; never silent, never from model knowledge). Mechanics in `guides.md` §0/§1; hooked from `SKILL.md` Operative Guides + Phase 4/5; reconciled with subagent dispatch in `dispatch.md`.
+- **Worktree/branch hygiene** in the closure discipline (`SKILL.md` Phase 4 isolate-on-branch / Phase 5 merge-decision + cleanup).
+- **Skill eval harness** (dev-only, not shipped): `test_skill_invariants.py` is the deterministic static release gate (`python -m unittest discover -s scripts -p "test_*.py"` — asserts the skill's own doctrine invariants: triggers/hook/worktree present and wired, indexes idempotent, support pointers resolve; zero LLM/network/subprocess). Opt-in behavioral corpus `evals/scenarios/` + `run_behavioral.py` (non-CI, never gates). `ENFORCEMENT.md` §5.
+
+### Process note
+- Full governance per unit: M-VISION v2.1 (revised — added the guide-consumption unit) → D-UC + P-TM → per-unit E-ISP (deep review) + E-TDD (light review) → implement → §4.6 code review. The independent-review gate caught real defects at design time (incomplete eval invariant set, the M3↔M4 dispatch interaction, a REPO path off-by-one, a P-TM overclaim of unbuilt guards). Battery 51/51 green. ADR `adr_2026-07-03_skill_eval_harness`; KL architecture v1.4 + principles v1.2.
+
+## [1.10.0] - 2026-07-03 (M3: Subagent Execution / Feature A — opt-in executable plan)
+### Added
+- **`dispatch.md`**: subagent-execution doctrine (opt-in for L3). The dispatch loop — validate the plan → per-task brief → economy-tier implementer → one-shot review → ledger — with client-relative model tiers (no provider names), one-shot review slots (no iterative loops), degradation to same-session where subagents do not exist, and guides injected by pointer (never pasted).
+- **`sdlc_check.py plan` subcommand**: `plan validate` (schema check of the executable plan, fail-closed path/guide confinement, sidecar-ledger cross-check — "no valid plan, no dispatch") and `plan brief --task <id>` (emits, to stdout, the task + prior-task interfaces + guide pointers). The validator is **zero-execution**: a task's `verify` command is emitted as text, never run.
+- **Executable-plan template** (`ai_docs/solutions/PLAN_[feature].md`) in `templates.md`: Markdown frontmatter (`status`, `derived-from`) + a fenced `json` task array + the sidecar `PLAN_[feature].ledger.json` shape (`task_id -> {status, verify_result, timestamp}`, git-tracked, survives compaction).
+- **`SKILL.md` §4** opt-in subagent-execution hook + the Hybrid `derived-from` seam (the plan is derived from the accepted E-TDD, never independently authored).
+- `test_plan.py`: stdlib-`unittest` battery for the `plan` subcommand (32 cases: schema, confinement, ledger, fail-fast JSON, zero-execution poka-yoke).
+
+### Changed
+- **`confine_under(base, rel)` extracted** in `sdlc_check.py`: the fail-closed path-confinement pattern (absolute/`..`/resolve-escape → reject), previously inlined twice (the `overrides:` and `distilled_from` checks), is now a single helper reused by both plus the new plan-path / guide-pointer confinement. Behavior-preserving (catches `(ValueError, OSError)`).
+
+### Process note
+- 4th live **model-per-dispatch** run (economy implementer from the E-TDD shadow, battery 32/32, deep code review PASS zero BLOCK). Governance: M-VISION → D-UC → P-TM → E-ISP → E-TDD, all through the independent review gate — which killed 3 real BLOCKs at design time (T1 subprocess-invariant misstatement, a missing impacted file, a `confine_under` OSError-crash regression). ADR `adr_2026-07-03_executable_plan_json_in_md`.
+
+## [1.9.0] - 2026-07-03 (M2 execution disciplines + Feature B unit 2 agent KB)
+### Added
+- `tdd.md`: TDD discipline (RED/GREEN/REFACTOR, increment rule, AAA test shape, documented exemptions) — the L2/L3 default for implementation work.
+- `debugging.md`: systematic debugging method (reproduce, isolate, root cause, fix, regression test, collateral check) with circuit-breaker integration.
+- `elicitation.md`: spec elicitation round (goal/benefit, scope boundaries, non-goals, constraints, acceptance signals) run before drafting the analysis in phase 3.
+- `review.md`: single definition of requesting, receiving, and performing code review, reused by the Hybrid review gates.
+- `SKILL.md` wiring: pointers to the four new files added at phase 3 (elicitation), phase 4 (TDD, debugging, circuit breaker), and phase 5 (review).
+- Agent-global KB (fixed root, project-wins precedence, `overrides:` with fail-closed confinement, collision warnings).
+
+## [1.8.1] - 2026-07-02
+### Fixed
+- **Guide freshness hash is now line-ending independent**: `sha256_file` in `sdlc_check.py` normalizes CRLF → LF before hashing. Previously the raw-byte hash made a fresh Windows checkout with `core.autocrlf=true` rewrite `.sources/` snapshots and flag every guide `[stale]` (false positive). Backward compatible: recorded hashes were computed on LF content, and normalization maps CRLF copies back to the same digest. (Edge case: a hash recorded pre-1.8.1 on a snapshot that genuinely contained CRLF bytes will flag `[stale]` once — regenerate the hash.)
+- `guides.md` step 3 now states the hash is computed over LF-normalized content and recommends the `ai_docs/reference/.sources/** -text` `.gitattributes` rule to consumer projects (defense in depth: keeps snapshots byte-verbatim).
+
+### Test battery addition
+- Scenario 10 (extends the unit-1 battery): snapshot checked out with CRLF endings + guide recording the LF-normalized hash → `stale` must NOT flag it; a genuine content edit must still flag `[stale]`.
+
+## [1.8.0] - 2026-07-02 (Feature B unit 1: operative guides, project scope)
+### Added
+- **Operative guides** (`ai_docs/reference/GUIDE_[topic].md`): a durable, source-faithful layer distilled from USER-PROVIDED indications — the capability neither agentic-sdlc nor superpowers had. New support file `guides.md` (pipeline: topic decomposition → user confirmation → verbatim snapshot in `reference/.sources/` with SHA-256 → source-anchored extraction → per-section fidelity markers `[source: …]` / `[not covered by source]`); guide template with provenance frontmatter (`source`, `source_version`, `distilled_from`, `source_hash`) in `templates.md`; short "Operative Guides" section in `SKILL.md`.
+- **Generated guide router** `ai_docs/reference/INDEX.md` (emitted by `sdlc_check.py index`, alignment-checked by `validate`): path, status, when-to-consult line and provenance summary per guide — the pointer target for the devPNT Hybrid bridge.
+- **Mechanical fidelity controls** in `sdlc_check.py`: provenance-key and per-section marker checks (warn; fails CI under `--strict`); `distilled_from` path confinement — absolute paths, `..` and symlink escapes rejected, fail-closed (ERROR); guide freshness in `stale` — recorded `source_hash` vs current snapshot, flagged in EVERY mode including `--hybrid` (guides are filesystem-first even in Hybrid).
+
+### Changed
+- `stale --hybrid` no longer returns unconditionally 0: it still skips audit-plan staleness (delegated to devPNT/KL) but now checks guide-source drift.
+- `list_canonical_docs` skips dot-subdirectories of canonical dirs (e.g. `reference/.sources/`): snapshot files are no longer swept into the manifest. Projects that kept `.md` files under dot-subdirs of canonical dirs will see them leave `INDEX.md` (more correct).
+
+### Process note
+- First live run of **model-per-dispatch**: implementation dispatched to an economy-tier subagent working from the accepted E-TDD shadow as a self-contained brief; independent deep code review passed first round with zero blocking findings. Governance: M-VISION → D-UC → P-TM → E-ISP → E-TDD, all through independent review gates (devPNT Hybrid).
+
+## [1.7.0] - 2026-07-02 (Phases 0-1 of the evolution roadmap)
+### Changed (breaking-soft)
+- **English is now the canonical language of the skill**: `SKILL.md`, `templates.md`, `ENFORCEMENT.md`, validator messages, generated indexes and the project protocol are in English. New ANALYSIS documents use English frontmatter keys (`status`, `level`, `start_date`, `end_date`) and English section headings. **Existing projects keep working**: the validator silently accepts the deprecated Italian keys (`stato`, `livello`, `data_inizio`, `data_fine`) and Italian headings.
+- The generated project protocol (`CLAUDE.md`/`GEMINI.md`/`AGENTS.md`/`.cursorrules`) is now a **thin pointer** to the skill (triage summary + where things live + closure gate) instead of a condensed copy of its rules, which had drifted from `SKILL.md`.
+- `init.js` now seeds `ai_docs/` from `templates.md` (single template source) instead of inline boilerplates; it creates `ai_docs/reference/` and the curated `ai_docs/README.md`, no longer seeds the generated `features_history.md`, and generates `INDEX.md` via the validator when Python is available, so the very first `sdlc_check.py check` on a fresh project is CLEAN.
+- Client detection unified between `init.js` and `postinstall.js` (`scripts/lib.js`): CLI on PATH **or** config home present (covers Claude Desktop with integrated Claude Code).
+
+### Added
+- `sdlc_check.py validate --strict` / `check --strict`: warnings and a missing `ai_docs/` become failures (for CI).
+- **Coexistence with devPNT (Phase 1, the Hybrid seam)**: new SKILL.md section with the ownership matrix (who is master per artifact in Standalone vs Hybrid), the triage equivalence table (one significance threshold, two vocabularies), the ANALYSIS↔plan-node state mapping and the shadow discipline (`SHADOW_[doc_key]_vX.Y.md`, exported BEFORE implementation; never saved under an `ANALYSIS_*` name).
+- `sdlc_check.py --hybrid` (explicit, never auto-detected) on `check`/`stale` (audit-plan staleness delegated to devPNT/KL) and on `gate` (an approved E-TDD shadow in `solutions/` authorizes writes on protected paths).
+- devPNT MCP doctrine (`mcp_system_prompt.md`, devPNT repo) slimmed accordingly: process (triage, phases, lifecycle, closure) deferred to the skill; `ai_docs/` layout aligned (adds `vision/` and `reference/`); shadow naming and shadow-before-implementation rule; checklist items for methodology, shadow export and hybrid closure gate.
+- Evolution roadmap for v1.7.0 in `ai_docs/vision/roadmap_evoluzione_agenti.md` (subagent execution, operative guides + agent-level knowledge base, devPNT seam, open-core positioning).
+
+### Fixed
+- Shadow detection is structural (filename `SHADOW_*` or `<!-- SHADOW` marker on the first line): an ANALYSIS merely *mentioning* shadows is no longer silently skipped from index and validation.
+- `mark`/`index` fail fast when `ai_docs/` is missing instead of silently creating a second documentation root in the wrong directory.
+
+### Removed
+- Divergent `agentic-sdlc-v2/` copy (integrated in 1.5.0, the leftover risked edits on the wrong files).
+- `references/*_template.md` (duplicated `templates.md` and had already diverged).
+
+## [1.6.0] - 2026-06-15
+### Added
+- **Manifest generato dei documenti canonici** (`ai_docs/INDEX.md`): `sdlc_check.py index` ora produce, oltre a `features_history.md`, un indice completo di tutti i doc in `vision/`, `reference/`, `architecture/`, `functional/`, `strategic/`, con descrizione e stato letti dall'header. Si rigenera, quindi non drifta.
+- **Lifecycle dei documenti canonici**: convenzione header `status: CURRENT|SUPERSEDED|DRAFT|DEPRECATED` + `supersedes:`. `validate` avvisa se `status` manca/è invalido o se un doc superseduto è ancora `CURRENT`. Stop ai grep che riportano a guide obsolete.
+- **Modello a due indici** documentato nella sezione "Documenti ai_docs" di `SKILL.md`: `README.md` curato (must-read, a mano) vs `INDEX.md` generato (completo, meccanico) — ruoli separati, prima confusi in un unico README che driftava.
+
+### Changed
+- §1 Audit: leggere `README.md` + `INDEX.md` all'avvio per sapere cosa esiste prima di esplorare il codice.
+- §3 Analisi: cercare con glob/grep un'ANALYSIS esistente prima di crearne una nuova (anti-duplicazione).
+- §5 Chiusura: gate "Indici allineati" — rigenerare `INDEX.md`, aggiornare il `README.md` curato per i must-read, marcare lo `status`; doc canonico non indicizzato o senza `status` = chiusura sporca.
+- `templates.md`: aggiunto il template dell'header dei documenti canonici.
+
+### Fixed
+- `sdlc_check.py` legge ora i file con `utf-8-sig`: un BOM iniziale (file autorati su Windows) non impedisce più il riconoscimento del frontmatter `---`.
+- L'estrattore dell'header riconosce sia il frontmatter `status:` sia la riga in corpo `**Status:**`/`Stato:`, e gli stati di tutte le convenzioni in uso (canonici `CURRENT/SUPERSEDED/DRAFT/DEPRECATED`, vision `DRAFT/APPROVED`, ADR `Accepted/Proposed/Rejected`) — niente più falsi avvisi "status non riconosciuto" su `APPROVED`/`Accepted`.
+- La descrizione del manifest salta righe di metadati (`Date`, `Created`, `Task ref`, ...) e i commenti HTML, così non finiscono come descrizione del documento.
+- `index` non genera più un `INDEX.md` vuoto su progetti senza documenti canonici (solo `solutions/`+`audit/`).
+
+### Migrazione (da 1.5.x)
+- Al primo `sdlc_check.py check`/`validate` dopo l'upgrade, un progetto con documenti canonici darà **un errore** `ai_docs/INDEX.md mancante`: è atteso — esegui **una volta** `sdlc_check.py index` per generarlo. Da lì in poi resta allineato.
+- I documenti canonici preesistenti senza `status:` produrranno **avvisi** (non errori): aggiungi l'header `description:`/`status:` per silenziarli. I nuovi progetti nascono già compatibili (template aggiornati).
+
+## [1.5.0] - 2026-06-13
+### Added
+- Introdotta la Regola Zero di triage (`L1`, `L2`, `L3`, `Spike`) per rendere il processo proporzionale al rischio.
+- Aggiunta simbiosi esplicita con devPNT: in Hybrid la `M-VISION` guida la milestone, il Master Plan resta roadmap strategica e l'Action Plan governa l'esecuzione tattica.
+- Aggiunti support file dentro la skill runtime: `templates.md`, `ENFORCEMENT.md`, `scripts/sdlc_check.py`.
+- `agentic-sdlc-install-skill` ora installa la skill nativa anche in `~/.gemini/skills/agentic-sdlc/`.
+- Aggiunto validatore meccanico opzionale per frontmatter ANALYSIS, Vision state, indice feature e audit stale.
+
+### Changed
+- Il nome pubblico resta `agentic-sdlc`; la proposta v2 e' stata integrata come evoluzione, non come skill parallela.
+- Aggiornati `agentic-sdlc-init`, template, protocolli generati, README e metadata.
+- La modalita Standalone resta completa; devPNT e' un livello di governance superiore, non un prerequisito.
+
+## [1.4.0] - 2026-06-07
+### Added
+- Introdotta la governance della **Vision** con nuova struttura `ai_docs/vision/` (`project_vision.md`, `roadmap.md`, `principles.md`, `features/`).
+- Aggiunto il **Vision Gate** nel workflow operativo: ogni feature significativa deve essere verificata rispetto a obiettivi, non-obiettivi, benefici attesi e segnali di successo prima dell'analisi tecnica.
+- Aggiunti template Vision in `references/` e sezione `Allineamento alla Vision` nel template di analisi.
+- `agentic-sdlc-init` ora crea i documenti Vision boilerplate nei nuovi progetti.
+
+## [1.3.1] - 2026-05-14
+### Fixed
+- Correzione documentazione (README + CHANGELOG) della sintassi per invocare il bin `agentic-sdlc-install-skill`. La forma `npx @antoneeo/agentic-sdlc-skill agentic-sdlc-install-skill` documentata in 1.3.0 **non funziona** perché npx non riesce a disambiguare il bin quando il pacchetto ne espone più di uno (errore: `could not determine executable to run`). Sintassi corretta: lanciare `agentic-sdlc-install-skill` direttamente dopo `npm install -g`, oppure usare `npx -p @antoneeo/agentic-sdlc-skill agentic-sdlc-install-skill` con `-p` esplicito.
+- Nessuna modifica al codice della skill: il bin di 1.3.0 funziona correttamente, era solo la doc a indicare la sintassi sbagliata.
+
+## [1.3.0] - 2026-05-14
+### Added
+- Nuovo comando esplicito `agentic-sdlc-install-skill` (registrato come `bin`): installa la skill in `~/.claude/skills/agentic-sdlc/` e `~/.codex/skills/agentic-sdlc/` **senza dipendere dal `postinstall` hook**. Risolve i casi in cui `npm` salta gli script (configurazioni `ignore-scripts=true`, policy IT aziendali, alcuni installer Node) e l'auto-install fallisce silenziosamente.
+- Sezione "Troubleshooting" nel README con istruzioni per il caso in cui la skill non venga rilevata da Claude Code dopo `npm install -g`.
+
+### Usage
+```bash
+npm install -g @antoneeo/agentic-sdlc-skill@latest
+agentic-sdlc-install-skill
+```
+
+## [1.2.4] - 2026-05-14
+### Fixed
+- `postinstall` ora rileva Claude Code anche quando il CLI `claude` non è nel PATH (es. Claude Desktop con Claude Code integrato): la presenza di `~/.claude/` o della variabile `CLAUDE_CONFIG_DIR` è sufficiente per attivare l'installazione della skill in `~/.claude/skills/agentic-sdlc/`. Stesso pattern già usato per Codex.
+- `preuninstall` rispetta `CLAUDE_CONFIG_DIR` per rimuovere la skill dalla directory di configurazione corretta.
+
+## [1.2.2] - 2026-05-10
+### Changed
+- README convertito in inglese per la pubblicazione npm.
+- Versione allineata a `1.2.2` in `package.json` e `gemini-extension.json`.
+
+## [1.2.1] - 2026-05-10
+### Fixed
+- `postinstall` ora registra la skill anche per Codex copiandola in `$CODEX_HOME/skills/agentic-sdlc/` oppure `~/.codex/skills/agentic-sdlc/`.
+- `preuninstall` rimuove anche la copia Codex della skill.
+- `agentic-sdlc-init` ora crea `AGENTS.md` per Codex invece di `.codex/hooks.json`, che Codex non carica come istruzioni di progetto.
+
+## [1.2.0] - 2026-05-10
+### Added
+- **Auto-installazione skill nativa Claude Code**: il `postinstall` ora copia `skills/agentic-sdlc-skill/` in `~/.claude/skills/agentic-sdlc/` quando rileva il CLI `claude`. La skill diventa disponibile come `agentic-sdlc` nel tool `Skill` di Claude Code dopo riavvio.
+- Nuovo script `preuninstall.js` che rimuove la skill da `~/.claude/skills/agentic-sdlc/` durante `npm uninstall`.
+- Fallback `copyRecursive` per Node < 16.7 (quando `fs.cpSync` non disponibile).
+
+### Changed
+- `package.json`: bump versione a 1.2.0, aggiunti keyword `claude-code` e `claude-skill`, descrizione aggiornata per riflettere supporto Claude Code nativo.
+
+### Fixed
+- Risolto bug per cui il pacchetto npm non registrava la skill in Claude Code (la cartella `~/.claude/skills/` non veniva mai popolata).
+
+## [1.1.0] - 2026-05-10
+### Fixed
+- Fixed path inconsistencies in the generated `CLAUDE.md`, `GEMINI.md`, and `.cursorrules` protocols. Added full paths (`ai_docs/strategic/`) to all document references to ensure AI agents (like Claude) can correctly find and update them.
+- Cleaned up encoding issues in initialization scripts.
+
+## [1.0.9] - 2026-05-10
+### Added
+- New **Smart Discovery** system during project initialization.
+- Official support for **Codex AI** via automatic `.codex/hooks.json` injection.
+- Support for **Cursor** and **Windsurf** via `.cursorrules`.
+- Binary command `agentic-sdlc-init` for use via `npx`.
+- `postinstall` script for global AI CLI detection (Claude, Gemini, Codex).
+- Full English localization for all user-facing messages and scripts.
+
+### Changed
+- Updated `init.js` to support the new `ai_docs/` directory structure defined in v1.0.8.
+
+## [1.0.8] - 2026-05-10
+### Aggiunto
+- Skill evoluta in "Hybrid Edition": integrazione opzionale con **devPNT** (server MCP) per governance avanzata tramite database e piani gerarchici.
+- Fase di Discovery per il rilevamento automatico dell'ambiente (Standalone vs Hybrid).
+- Supporto per ADR (Architecture Decision Records) e Knowledge Layer (KL) nella fase di chiusura.
+
+### Modificato
+- Riorganizzazione dei percorsi di documentazione (`ai_docs/strategic/`, `ai_docs/audit/`, `ai_docs/solutions/`).
+- Aggiornata la documentazione funzionale (`architecture_overview.md`, `external_interfaces.md`) con definizioni più precise.
+
+## [1.0.7] - 2026-05-10
+### Modificato
+- Aggiornata l'attribuzione dell'autore (Antonio Pinto) e il copyright in tutti i file (`package.json`, `README.md`, `SKILL.md`).
+- Aggiunto link al profilo GitHub ufficiale.
+
+## [1.0.6] - 2026-05-10
+### Aggiunto
+- Sincronizzazione completa dei file del progetto nel repository.
+- Supporto per il tracciamento delle feature e audit plan.
+
+## [1.0.5] - 2026-04-17
+### Aggiunto
+- File `CHANGELOG.md` per il tracciamento delle versioni.
+
+## [1.0.4] - 2026-04-17
+### Modificato
+- `README.md`: Aggiunte istruzioni specifiche per l'installazione da chiavetta USB e chiarimenti sulla disponibilità globale della skill.
+
+## [1.0.3] - 2026-04-17
+### Aggiunto
+- Primo `README.md` con istruzioni di installazione e attivazione.
+
+## [1.0.2] - 2026-04-17
+### Corretto
+- Riorganizzata la struttura delle cartelle secondo gli standard di Gemini CLI (`skills/agentic-sdlc-skill/SKILL.md`).
+- Aggiunto frontmatter YAML a `SKILL.md` per la scoperta automatica.
+
+## [1.0.1] - 2026-04-17
+### Corretto
+- Aggiunto `gemini-extension.json` (manifest dell'estensione) mancante nella versione iniziale.
+
+## [1.0.0] - 2026-04-17
+### Iniziale
+- Prima pubblicazione della skill (Protocollo SDLC Documentation-First).