@tacuchi/agent-workflow-cli 17.1.0 → 19.0.0

package/skills/w/commands/fix-git.md

@@ -17,7 +17,7 @@ Single-pass, **no loop, no session**, **never writes `docs/`**. **Transversal**
 
 1. **Detect + identify** — run `aw merge-state [<source>]` (read-only; `--source <alias>` or `--all` when a workspace exists; a direct path otherwise). From the JSON, per repo: `is_merging`, `current_branch` (**destination / ours**), `merge_origin` (**origin / theirs**), `conflicted_files`.
    - If **no merge is in progress** (`is_merging:false`) and the user named a **target** (e.g. "merge `<branch>`"): that is an explicit request → `git -C <path> merge <branch>` and continue. No target → report there is no merge to resolve and stop.
-2. **Resolve** — **read and follow** the ***Merge-conflict resolution*** section of the `git` role (`../roles/git/SKILL.md`): analyze each conflict's intent (3 versions `git show :1:/:2:/:3:<file>`, `git log --merge`), resolve (ours / theirs / combine / rewrite) and `git add` what is resolved. On **ambiguity or incoherence**, ask via *structured-choice* (never invent the resolution).
+2. **Resolve** — **read and follow** the ***Merge-conflict resolution*** section of the `git` role (`../roles/git/ROLE.md`): analyze each conflict's intent (3 versions `git show :1:/:2:/:3:<file>`, `git log --merge`), resolve (ours / theirs / combine / rewrite) and `git add` what is resolved. On **ambiguity or incoherence**, ask via *structured-choice* (never invent the resolution).
 3. **Close** — **propose** the merge commit (propose-then-execute, canonical format, git-safe). Escape hatch: `git merge --abort` after user confirmation.
 
 > Do not try `Skill: git` — the role is **read and followed** (it is the capability this command composes). The command **is** the entry; the conflict doctrine lives in the `git` role.
@@ -28,6 +28,6 @@ Run `aw merge-state` (read-only), report **origin ↔ destination** and the per-
 
 ## Resources
 
-- Capability: `../roles/git/SKILL.md` (section *Merge-conflict resolution*)
+- Capability: `../roles/git/ROLE.md` (section *Merge-conflict resolution*)
 - CLI: `aw merge-state` (read-only merge-state inspector)
 - Design reference: `docs/referencias/workflow-skills/fix-git.md`

package/skills/w/commands/plan-exec.md

@@ -25,7 +25,7 @@ Starts or resumes `plan-exec-loop` (Layer 2), which executes the real work phase
 
 `plan-exec-loop` is **not** a skill invocable by name — it is this command's operating manual (a sibling doc in the bundle). **Load it and execute it end to end**:
 
-1. **Read** `../loops/plan-exec-loop/SKILL.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
+1. **Read** `../loops/plan-exec-loop/LOOP.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
 2. **Follow** its instructions taking `$ARGUMENTS` as input: it detects CHECKPOINT/resume, executes phase by phase (git-safe, DB scripts-only), keeps the plan alive and reports.
 
 > Do not try `Skill: plan-exec-loop` — it is not registered as a skill. The command **is** the entry; the loop is its body.
@@ -35,7 +35,7 @@ Starts or resumes `plan-exec-loop` (Layer 2), which executes the real work phase
 - Reads and updates `docs/plans/PPP-plan-<slug>.md` (living doc: phase/task state).
 - Edits code in the workspace sources (a single execution session per run; execution is still phase by phase, there is just no session per phase).
 - If it creates a tool/utility, the ambient `creating-tools` skill documents it in `docs/tools/` (auto-discovered; the workflow does not bind it).
-- **Closing review gate** at every phase boundary, **before proposing the commits**: re-reads the diff (independent pass) applying the **installed ambient conventions** and fixes or defers findings — nothing reaches a commit unreviewed (see `../loops/plan-exec-loop/SKILL.md` § *Delta 5*).
+- **Closing review gate** at every phase boundary, **before proposing the commits**: re-reads the diff (independent pass) applying the **installed ambient conventions** and fixes or defers findings — nothing reaches a commit unreviewed (see `../loops/plan-exec-loop/LOOP.md` § *Delta 5*).
 - Proposes commits per source (git-safe: verifies the branch, proposes, never push/--amend/--no-verify).
 - Generates session artifacts (`DECISION`, `SCRIPTS.sql`) under `.workflow/sessions/`.
 - **Never exports** to `docs/scripts`, `docs/manuals`, `docs/diagrams`, `docs/reports` — the `export-*` do that as a separate step.
@@ -51,5 +51,5 @@ The skill describes, phase by phase, what it would execute, which files it would
 
 ## Resources
 
-- Loop skill: `../loops/plan-exec-loop/SKILL.md`
+- Loop manual: `../loops/plan-exec-loop/LOOP.md`
 - Design reference: `docs/referencias/workflow-commands/plan-exec.md`

package/skills/w/commands/plan-new.md

@@ -35,7 +35,7 @@ The skill evaluates `$ARGUMENTS` (specs live in place — `docs/specs/NNN-spec-<
 
 `plan-new-loop` is **not** a skill invocable by name — it is this command's operating manual (a sibling doc in the bundle). **Load it and execute it end to end**:
 
-1. **Read** `../loops/plan-new-loop/SKILL.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
+1. **Read** `../loops/plan-new-loop/LOOP.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
 2. **Follow** its instructions taking `$ARGUMENTS` as input (resolved per the 3 rules above): it detects state/resume, runs the gap-driven engine, creates and manages sessions, converges and reports.
 
 > Do not try `Skill: plan-new-loop` — it is not registered as a skill. The command **is** the entry; the loop is its body.
@@ -46,7 +46,7 @@ The plan is named `docs/plans/PPP-plan-<slug>.md`. `aw next-number docs/plans` r
 
 ## UI → design SPECs
 
-If the plan **includes UI**, the loop composes the `ui-design` capability and produces per-screen **design SPECs** (`NNN-SPEC-<SLUG>.md`) as artifacts of its session — the plan's UI Tasks reference them (see `../loops/plan-new-loop/SKILL.md` § *Delta 4* and `../artifacts/artifacts-design/SPEC.md`).
+If the plan **includes UI**, the loop composes the `ui-design` capability and produces per-screen **design SPECs** (`NNN-SPEC-<SLUG>.md`) as artifacts of its session — the plan's UI Tasks reference them (see `../loops/plan-new-loop/LOOP.md` § *Delta 4* and `../artifacts/artifacts-design/SPEC.md`).
 
 ## Plan mode
 
@@ -54,5 +54,5 @@ The skill resolves the input per the 3 rules above and describes the loop action
 
 ## Resources
 
-- Loop skill: `../loops/plan-new-loop/SKILL.md`
+- Loop manual: `../loops/plan-new-loop/LOOP.md`
 - Design reference: `docs/referencias/workflow-commands/plan-new.md`

package/skills/w/commands/plan-refine.md

@@ -36,7 +36,7 @@ The skill evaluates `$ARGUMENTS` (plans live in place — `docs/plans/PPP-plan-<
 
 `plan-refine-loop` is **not** a skill invocable by name — it is this command's operating manual (a sibling doc in the bundle). **Load it and execute it end to end**:
 
-1. **Read** `../loops/plan-refine-loop/SKILL.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
+1. **Read** `../loops/plan-refine-loop/LOOP.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
 2. **Follow** its instructions taking `$ARGUMENTS` as input: it detects state/resume, runs the gap-driven engine, creates and manages sessions, converges and reports.
 
 > Do not try `Skill: plan-refine-loop` — it is not registered as a skill. The command **is** the entry; the loop is its body.
@@ -52,7 +52,7 @@ The skill detects prior state before starting, **keying off the `CHECKPOINT`** (
 
 ## UI → design SPECs
 
-If the refine **touches UI**, the loop composes `ui-design` and produces/updates **design SPECs** (`NNN-SPEC-<SLUG>.md`) in its own session — scoped to the new/changed screens — and re-points the plan references (see `../loops/plan-refine-loop/SKILL.md` § *Delta 4*).
+If the refine **touches UI**, the loop composes `ui-design` and produces/updates **design SPECs** (`NNN-SPEC-<SLUG>.md`) in its own session — scoped to the new/changed screens — and re-points the plan references (see `../loops/plan-refine-loop/LOOP.md` § *Delta 4*).
 
 ## Plan mode
 
@@ -60,5 +60,5 @@ The skill resolves the state and describes the actions the loop would run (gaps
 
 ## Resources
 
-- Loop skill: `../loops/plan-refine-loop/SKILL.md`
+- Loop manual: `../loops/plan-refine-loop/LOOP.md`
 - Design reference: `docs/referencias/workflow-commands/plan-refine.md`

package/skills/w/commands/quick.md

@@ -25,7 +25,7 @@ For scoped, direct tasks that do not justify going through SPEC or PLAN. Creates
 
 `quick-loop` is **not** a skill invocable by name — it is this command's operating manual (a sibling doc in the bundle). **Load it and execute it end to end**:
 
-1. **Read** `../loops/quick-loop/SKILL.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
+1. **Read** `../loops/quick-loop/LOOP.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
 2. **Follow** its instructions taking `$ARGUMENTS` as the task: it evaluates the size gate, creates the light session, works with minimal ceremony (git-safe), escalates if the task exceeds or grows (SPEC live / PLAN deferred), and reports.
 
 > Do not try `Skill: quick-loop` — it is not registered as a skill. The command **is** the entry; the loop is its body.
@@ -34,9 +34,9 @@ For scoped, direct tasks that do not justify going through SPEC or PLAN. Creates
 
 - Edits code in the workspace sources.
 - Minimal session artifacts (lazy DECISION, proposed commit).
-- **Proportional closing review gate** before proposing the single commit: re-reads the diff applying the installed ambient conventions and fixes or defers (see `../loops/quick-loop/SKILL.md` § *Sequence*).
+- **Proportional closing review gate** before proposing the single commit: re-reads the diff applying the installed ambient conventions and fixes or defers (see `../loops/quick-loop/LOOP.md` § *Sequence*).
 - **Never touches `docs/`** and exports nothing.
-- **Escalates** when complexity emerges — **entry size gate** (before creating the session) and mid-loop (many files, ≥2 sources, needs architecture, or the change is a feature/refactor). Accepting **SPEC** = **live** transition into the SPEC flow (draft via the spec-new procedure + spec-refine-loop); **PLAN** stays seeded for later. See `../loops/quick-loop/SKILL.md` § *QUICK delta*.
+- **Escalates** when complexity emerges — **entry size gate** (before creating the session) and mid-loop (many files, ≥2 sources, needs architecture, or the change is a feature/refactor). Accepting **SPEC** = **live** transition into the SPEC flow (draft via the spec-new procedure + spec-refine-loop); **PLAN** stays seeded for later. See `../loops/quick-loop/LOOP.md` § *QUICK delta*.
 
 ## Plan mode
 
@@ -44,5 +44,5 @@ The skill describes the changes it would apply and the files it would touch, wit
 
 ## Resources
 
-- Loop skill: `../loops/quick-loop/SKILL.md`
+- Loop manual: `../loops/quick-loop/LOOP.md`
 - Design reference: `docs/referencias/workflow-commands/quick.md`

package/skills/w/commands/spec-new.md

@@ -57,7 +57,7 @@ Assumed facts.
 Pending doubts. ← the spec-refine-loop closes them.
 ```
 
-> **`Open questions` goes last** — the refined spec **inserts before `Open questions`** `## UI spec` (if there is UI) + `## Refinement decisions` + `## Q&A traceability` (refined schema in the [`spec-refine-loop`](../loops/spec-refine-loop/SKILL.md)). Same skeleton: the draft and the refined spec share the order.
+> **`Open questions` goes last** — the refined spec **inserts before `Open questions`** `## UI spec` (if there is UI) + `## Refinement decisions` + `## Q&A traceability` (refined schema in the [`spec-refine-loop`](../loops/spec-refine-loop/LOOP.md)). Same skeleton: the draft and the refined spec share the order.
 
 **Filling notes:**
 
@@ -69,7 +69,7 @@ Pending doubts. ← the spec-refine-loop closes them.
 - The **gaps** the loop detects = weak sections of the schema (vague Requirement, Scope without `Out`, untestable criteria, open questions, undeclared assumptions, contradictions) **+ UI unspecified** when the requirement involves UI.
 - Equivalent alternative: the user creates the draft by hand. Both paths produce the same `docs/specs/NNN-spec-<slug>.md`.
 
-> **Reuse by escalation:** the live escalation from `/w:quick` (see [`../loops/quick-loop/SKILL.md`](../loops/quick-loop/SKILL.md) § *QUICK delta*) materializes its draft following **this same procedure** (steps 1-3: same schema, same NO RESEARCH single-pass hard rule), with `## Origin` = "escalated from `/w:quick`" + the original prompt. No need to type `/w:spec-new`: the consent in the structured-choice equals invoking it.
+> **Reuse by escalation:** the live escalation from `/w:quick` (see [`../loops/quick-loop/LOOP.md`](../loops/quick-loop/LOOP.md) § *QUICK delta*) materializes its draft following **this same procedure** (steps 1-3: same schema, same NO RESEARCH single-pass hard rule), with `## Origin` = "escalated from `/w:quick`" + the original prompt. No need to type `/w:spec-new`: the consent in the structured-choice equals invoking it.
 
 ## Plan mode
 
@@ -78,4 +78,4 @@ Resolves `NNN` by reading `docs/specs/`, describes the draft it would generate w
 ## Resources
 
 - Design reference: `docs/referencias/workflow-commands/spec-new.md`
-- Loop that refines this draft: `../loops/spec-refine-loop/SKILL.md`
+- Loop that refines this draft: `../loops/spec-refine-loop/LOOP.md`

package/skills/w/commands/spec-refine.md

@@ -25,7 +25,7 @@ This command does not refine the spec itself: it delegates to `spec-refine-loop`
 
 `spec-refine-loop` is **not** a skill invocable by name — it is this command's operating manual (a sibling doc in the bundle). **Load it and execute it end to end**:
 
-1. **Read** `../loops/spec-refine-loop/SKILL.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
+1. **Read** `../loops/spec-refine-loop/LOOP.md` (inside the installed `w` skill — e.g. `~/.claude/skills/w/loops/…`).
 2. **Follow** its instructions taking `$ARGUMENTS` as input: it detects state/resume, runs the gap-driven engine, creates and manages sessions, converges and reports.
 
 > Do not try `Skill: spec-refine-loop` — it is not registered as a skill. The command **is** the entry; the loop is its body.
@@ -47,5 +47,5 @@ The skill resolves the state and describes the actions the loop would run (gaps
 
 ## Resources
 
-- Loop skill: `../loops/spec-refine-loop/SKILL.md`
+- Loop manual: `../loops/spec-refine-loop/LOOP.md`
 - Design reference: `docs/referencias/workflow-commands/spec-refine.md`

package/skills/w/commands/workspace-init.md

@@ -1,5 +1,5 @@
 ---
-description: Turns the current folder into an agent-workflow workspace (scaffolding .workflow/ + docs/ + WORKSPACE block + skills.toml). Replaces hub-init + project-init — no project/hub distinction. Run once before any flow; idempotent.
+description: Turns the current folder into an agent-workflow workspace (minimal scaffold — .workflow/sessions + skills.toml + WORKSPACE block + CLI-owned .gitignore; docs/ folders are born on demand). Replaces hub-init + project-init — no project/hub distinction. Run once before any flow; idempotent, re-running reconciles (prunes legacy empty scaffold).
 argument-hint: --source alias:path[:branch] [--proyecto <name>] [--main-branch <branch>] [--dry-run]
 allowed-tools:
   [
@@ -22,16 +22,18 @@ aw workspace-init --source alias:path[:branch] [--proyecto <name>] [--main-branc
 
 1. **Detect/confirm sources** — the CLI detects the repo path(s); the user confirms aliases, paths and branches. Multiple `--source` accepted.
 2. **Pick default skills** — the catalog of available capabilities (roles) is presented. Per role: `built-in default`, override to a third-party skill (`skills.sh`), or `off`. The result is written to `.workflow/skills.toml`. Config cascade: `built-in → ~/.workflow/skills.toml (global) → .workflow/skills.toml (workspace)`.
-3. **Write the scaffolding** — creates `.workflow/sessions/`, `docs/` with its taxonomy (`specs/`, `plans/`, `scripts/`, `manuals/`, `diagrams/`, `reports/`), the `WORKSPACE` block in CLAUDE.md/AGENTS.md (sources + metadata), and `.workflow/skills.toml`.
-4. **Multi-source** — with ≥2 sources, configures multi-root visibility (settings.local.json + config, gitignored) and reconciles sources.
+3. **Write the MINIMAL scaffolding** — only the activation set: `.workflow/sessions/` (the marker that activates the operating context), `.workflow/skills.toml`, the `WORKSPACE` block in CLAUDE.md/AGENTS.md (sources + metadata) and the `.gitignore` policy (below). **Nothing else upfront**: each `docs/<category>` folder (`specs/`, `plans/`, `scripts/`, `manuals/`, `diagrams/`, `reports/`) is born on demand at its first numbered write — `aw next-number docs/<cat>` creates it (skills stay unchanged); launch artifacts (`.workflow/launch/<alias>/`) and `docs/logs/` are generated by the launch flow at the first launch.
+4. **External sources** — for every source living OUTSIDE the workspace folder (any count, including a single external source), configures multi-root visibility (settings.local.json + config, gitignored) and reconciles sources.
+
+**Versioning policy (CLI-owned `.gitignore`)** — init writes and owns the full set: ignored = `.workflow/sessions/` (machine-local live log), `.workflow/.lock`, `.workflow/processes.json`, `.workflow/launch/`, `docs/logs/`, and — with external sources — `.claude/settings.local.json*` / `.codex/config.toml*` (patterns cover the `.bak.<epoch>` backups). Versioned = `.workflow/skills.toml`, `docs/**` (deliverables) and `.workflow/HISTORY.md` — the durable record: `aw session-close` upserts each closed session's row there.
 
 When done, the user can run `/w:spec-new`, `/w:plan-new` or `/w:quick` directly.
 
-**Idempotent**: re-running reconciles (no duplicate entries, no overwriting manual configuration).
+**Idempotent + reconcile**: re-running reconciles (no duplicate entries, no overwriting manual configuration) and **prunes the legacy upfront scaffold**: `.gitkeep`-only taxonomy folders, stray `.gitkeep` files, an empty `docs/logs/` and a released `.workflow/.lock` leftover are removed; the `.gitignore` is completed to the current set (entries merge under the existing header, never duplicated).
 
 ## Plan mode
 
-Resolves the sources and describes the scaffolding it would create, without writing files. Shows what it would create under `.workflow/` and `docs/`, and what it would write into CLAUDE.md.
+Resolves the sources and describes the minimal scaffolding it would create (and what a re-run would prune), without writing files.
 
 ## Resources
 

package/skills/w/exports/README.md

@@ -3,7 +3,7 @@
 > This is the **bundle README** for the `export-*` family: the **only** path that promotes session artifacts to permanent `docs/` documents. Each export is invoked by the **user** (never by a loop) as a separate, explicit step.
 > Related layers: [`../commands/`](../commands/) (Layer 1 flows) · [`../loops/`](../loops/) (Layer 2, AI-driven) · artifacts live in `.workflow/sessions/` (Layer 3). Design reference: `docs/referencias/workflow-exports/`.
 >
-> **Namespace:** these are skills registered **by name** (`name:` — `export-scripts`, `export-manuals`, `export-diagrams`, `export-reports`); in Claude Code they are reached via the `Skill` tool — the by-name binding of the procedure-loading capability, not a universal (see [`../harness/SKILL.md`](../harness/SKILL.md)). The thin `/w:export-*` slash commands that route here are authored separately under [`../commands/`](../commands/).
+> **Namespace:** each export body is an **operating manual** (`EXPORT.md` — deliberately not a `SKILL.md`, so no host indexes it as a standalone skill). The user-invocable surface is the `/w:export-*` commands authored under [`../commands/`](../commands/), which read-and-follow the sibling `EXPORT.md` (per-host wrapper: see [`../harness/HARNESS.md`](../harness/HARNESS.md) § *Command packaging*).
 
 ---
 
@@ -24,10 +24,10 @@
 
 | Export | Composes | Reads (artifacts / sessions + corpus) | Writes (its ONLY category) |
 |---|---|---|---|
-| [`export-scripts`](export-scripts/SKILL.md) | `sql` | type-B `SCRIPTS.sql` (DDL/DML migrations) across N sessions + standalone `docs/scripts/*.sql` | `docs/scripts/NNN-export-scripts-<date>/` (numbered forwards + `00-ROLLBACK.sql`) |
-| [`export-manuals`](export-manuals/SKILL.md) | — (prose: ambient conventions) | sessions + `DECISION` + plan-doc (`Solution`, `Final behavior`, `Validations`) + touched code | `docs/manuals/` |
-| [`export-diagrams`](export-diagrams/SKILL.md) | `diagrams` | source code of the sources + plan-doc (`AS-IS` / `TO-BE`, `Impacted`) | `docs/diagrams/` (C4 / mermaid) |
-| [`export-reports`](export-reports/SKILL.md) | — (prose: ambient conventions) | corpus of sessions (spec, `CONCLUSIONS`, `DECISION`) + plan-doc state + `docs/` | `docs/reports/` (executive / functional report) |
+| [`export-scripts`](export-scripts/EXPORT.md) | `sql` | type-B `SCRIPTS.sql` (DDL/DML migrations) across N sessions + standalone `docs/scripts/*.sql` | `docs/scripts/NNN-export-scripts-<date>/` (numbered forwards + `00-ROLLBACK.sql`) |
+| [`export-manuals`](export-manuals/EXPORT.md) | — (prose: ambient conventions) | sessions + `DECISION` + plan-doc (`Solution`, `Final behavior`, `Validations`) + touched code | `docs/manuals/` |
+| [`export-diagrams`](export-diagrams/EXPORT.md) | `diagrams` | source code of the sources + plan-doc (`AS-IS` / `TO-BE`, `Impacted`) | `docs/diagrams/` (C4 / mermaid) |
+| [`export-reports`](export-reports/EXPORT.md) | — (prose: ambient conventions) | corpus of sessions (spec, `CONCLUSIONS`, `DECISION`) + plan-doc state + `docs/` | `docs/reports/` (executive / functional report) |
 
 > **Composition over ownership:** an export that owns a derived artifact does **not** own its authoring logic — it **composes a capability role** from [`../roles/`](../roles/) (resolved through `.workflow/skills.toml`): `export-scripts` composes `sql`; `export-diagrams` composes `diagrams`. Swapping the implementation is a one-line config change; it never touches the export. `export-manuals` and `export-reports` produce **prose**, which follows **ambient writing conventions** (the host auto-applies an installed writing skill if present) — they do **not** compose or bind a `writing` role.
 
@@ -36,11 +36,11 @@
 1. **Layer 1, explicit** — the **user** invokes them (`/w:export-<cat>`). **Never** automatic (no loop fires them).
 2. **Single-pass, read-only over sessions** — they read artifacts/sessions and `docs/`, **synthesize**, and write **only** their own `docs/<category>/` folder. They do **not** mutate sessions and do **not** open/close loops.
 3. **Cross-session** — they consolidate **N** sessions + the `docs/` corpus (dedup, roadmap, continuous numbering).
-4. **No loop, no internal sessions** — options come from **args** (no lifecycle *structured-choice*; harness capability — see [`../harness/SKILL.md`](../harness/SKILL.md)).
+4. **No loop, no internal sessions** — options come from **args** (no lifecycle *structured-choice*; harness capability — see [`../harness/HARNESS.md`](../harness/HARNESS.md)).
 5. **Git-safe** — they **never** commit, merge, push, `--amend`, or `--no-verify`. The output is a written document the user reviews and commits when ready.
 6. **DB scripts-only** — `export-scripts` ships migration SCRIPTS as a bundle; it **never executes** DDL/DML (a human/DBA applies them).
 
-## Section schema of each `export-*/SKILL.md`
+## Section schema of each `export-*/EXPORT.md`
 
 Mirrors `docs/referencias/workflow-exports/` and the old export SKILLs. Frontmatter: `name:` (kebab — exactly `export-scripts` / `export-manuals` / `export-diagrams` / `export-reports`) + rich `description:` (what + when, drives selection). Body:
 
@@ -63,11 +63,11 @@ Mirrors `docs/referencias/workflow-exports/` and the old export SKILLs. Frontmat
 Exports read the corpus through the CLI — **never hard-coded paths**:
 
 - `aw sessions` — list sessions (counts + next correlative) to enumerate the corpus.
-- `aw release-data [--since sessionNNN] [--source alias]` — consolidated dump of sessions (corpus enumeration).
+- `aw release-data [--since sessionNNN] [--source alias] [--include-graduated] [--standalone-sql]` — consolidated dump of sessions (corpus enumeration). `--include-graduated` lists previous `docs/scripts` bundles (modern `NNN-export-scripts-YYYY-MM-DD` and legacy naming); `--standalone-sql` lists loose top-level `docs/scripts/*.sql` (export-scripts' source B).
 - `aw session-artifacts --code <NNN> [--dump [kinds]]` — counts by default; `--dump` returns `{path, content, size}` per artifact (objetivo, decisiones, conclusiones, tasks, checkpoint, backlog, scripts).
-- `aw next-number docs/<category>` — deterministic numbering of the output.
+- `aw next-number docs/<category>` — deterministic numbering of the output. It also **creates the category folder when missing** (workspace-init no longer scaffolds docs/ upfront) — this is what makes destination resolution a CLI guarantee. In plan/dry-run mode call it with `--dry-run` (pure query, creates nothing).
 
-> The destination-folder resolution (workspace root, single- vs multi-source) is handled by the CLI internally. If a specific flag is uncertain at implementation time, it is noted inline in each SKILL.
+> The destination-folder resolution (workspace root, single- vs multi-source, on-demand creation) is handled by the CLI internally. If a specific flag is uncertain at implementation time, it is noted inline in each SKILL.
 
 ## 6 Hard invariants (never violate)
 
@@ -82,7 +82,7 @@ Exports read the corpus through the CLI — **never hard-coded paths**:
 
 | Export | File | Category | Composes |
 |---|---|---|---|
-| `export-scripts` | [`export-scripts/SKILL.md`](export-scripts/SKILL.md) | `docs/scripts` | `sql` |
-| `export-manuals` | [`export-manuals/SKILL.md`](export-manuals/SKILL.md) | `docs/manuals` | — (prose: ambient conventions) |
-| `export-diagrams` | [`export-diagrams/SKILL.md`](export-diagrams/SKILL.md) | `docs/diagrams` | `diagrams` |
-| `export-reports` | [`export-reports/SKILL.md`](export-reports/SKILL.md) | `docs/reports` | — (prose: ambient conventions) |
+| `export-scripts` | [`export-scripts/EXPORT.md`](export-scripts/EXPORT.md) | `docs/scripts` | `sql` |
+| `export-manuals` | [`export-manuals/EXPORT.md`](export-manuals/EXPORT.md) | `docs/manuals` | — (prose: ambient conventions) |
+| `export-diagrams` | [`export-diagrams/EXPORT.md`](export-diagrams/EXPORT.md) | `docs/diagrams` | `diagrams` |
+| `export-reports` | [`export-reports/EXPORT.md`](export-reports/EXPORT.md) | `docs/reports` | — (prose: ambient conventions) |

package/skills/w/exports/export-diagrams/{SKILL.md → EXPORT.md}

@@ -43,13 +43,13 @@ The **`diagrams`** capability (built-in default `diagrams`), resolved via `.work
 
 ## Read-only sandbox
 
-In plan mode it **describes**, never writes: the resolved engine, the levels/sections that would appear (resolved by args), the sources to inspect + detected integrations, and — with a data-model request — the proposed MCP queries with their estimated cost. It does **not** run `Write`, MCP mutations, or effectful `aw next-number`.
+In plan mode it **describes**, never writes: the resolved engine, the levels/sections that would appear (resolved by args), the sources to inspect + detected integrations, and — with a data-model request — the proposed MCP queries with their estimated cost. It does **not** run `Write` or MCP mutations; numbering queries use `aw next-number --dry-run` (pure).
 
 ## Inputs
 
 **`agent-workflow` CLI (alias `aw`)** — never read hardcoded paths:
 
-- `aw sessions` / `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates the corpus (input for the AS-IS/TO-BE delta).
+- `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates the corpus (ALL sessions; input for the AS-IS/TO-BE delta). `aw sessions` alone lists only ACTIVE sessions — never use it as the corpus.
 - `aw session-artifacts --code <NNN> --dump objetivo` — locates the session and its plan-doc reference; `AS-IS`/`TO-BE`/`Impacted` are read from the plan-doc by its path.
 - `aw next-number docs/diagrams` — deterministic numbering (the CLI handles destination-folder resolution).
 
@@ -60,7 +60,7 @@ In plan mode it **describes**, never writes: the resolved engine, the levels/sec
 
 **Read-only MCP** (optional, only with a data-model request and configuration): `\d <table>`, `SELECT count(*)`, FK relations for the `erDiagram`. With the cost guard.
 
-**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/SKILL.md`](../../harness/SKILL.md)):
+**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/HARNESS.md`](../../harness/HARNESS.md)):
 
 ```
 /w:export-diagrams [--sessions NNN[,NNN]] [--since sessionNNN] [--source <alias>]
@@ -82,7 +82,7 @@ No args: `--engine mermaid --scope todo`. The system **snapshot** is always the
 
 ### Step 1 — Resolve context and corpus
 
-`aw sessions` / `release-data` applying `--sessions`/`--since`/`--source`. The CLI handles destination-folder resolution.
+`aw release-data` applying `--sessions`/`--since`/`--source`. The CLI handles destination-folder resolution.
 
 ### Step 2 — Inspect the sources
 
@@ -98,7 +98,7 @@ If `--scope` includes `data` and read-only MCP exists: `\d <table>`, `count(*)`,
 
 ### Step 5 — Render (composes `diagrams`)
 
-Per `--engine`: `mermaid` → native Mermaid C4 blocks (`C4Context`/`C4Container`/`C4Component`) and `flowchart` for flows; `c4` → a separate Structurizr `workspace.dsl` + auxiliary embedded Mermaid for offline reading. For every ```` ```mermaid ```` block, add immediately after the closing fence a blockquote with the preview link: `> Ver diagrama renderizado: <https://mermaid.ink/img/BASE64>` (URL-safe base64 of the plain code). Not applicable to `workspace.dsl`.
+Per `--engine`: `mermaid` → native Mermaid C4 blocks (`C4Context`/`C4Container`/`C4Component`) and `flowchart` for flows; `c4` → a separate Structurizr `workspace.dsl` + auxiliary embedded Mermaid for offline reading. Keep block labels ASCII (no accents) for render robustness; prose keeps them. OPTIONALLY add after each closing fence a preview blockquote `> Ver diagrama renderizado: <https://mermaid.ink/img/BASE64>` (URL-safe base64, keep padding) — the link encodes the diagram source into a public-service URL, omit it for private corpora (GitHub renders the blocks natively). Not applicable to `workspace.dsl`.
 
 ### Step 6 — Write or report
 
@@ -122,4 +122,4 @@ Functionally idempotent: each invocation takes the next `NNN`; it never overwrit
 - Design: `docs/referencias/workflow-exports/export-diagrams.md` · family: [`../README.md`](../README.md).
 - Composed capability: `diagrams` (built-in default; see `docs/referencias/workflow-roles/`).
 - Input: plan-doc `AS-IS`/`TO-BE`/`Impacted` (see `docs/plans`).
-- Siblings: [`../export-scripts/SKILL.md`](../export-scripts/SKILL.md) · [`../export-manuals/SKILL.md`](../export-manuals/SKILL.md) · [`../export-reports/SKILL.md`](../export-reports/SKILL.md).
+- Siblings: [`../export-scripts/EXPORT.md`](../export-scripts/EXPORT.md) · [`../export-manuals/EXPORT.md`](../export-manuals/EXPORT.md) · [`../export-reports/EXPORT.md`](../export-reports/EXPORT.md).

package/skills/w/exports/export-manuals/{SKILL.md → EXPORT.md}

@@ -44,13 +44,13 @@ The manual's prose follows the **ambient** writing conventions: the host auto-ap
 
 ## Read-only sandbox
 
-In plan mode it **describes**, never writes: the resolved mode, the detected topics (with origin sessions), the manuals already present in `docs/manuals/`, and — per mode — the `INDEX.md` structure it would overwrite or the count of manuals the dossier would generate. It does **not** run `Write` or effectful `aw next-number`.
+In plan mode it **describes**, never writes: the resolved mode, the detected topics (with origin sessions), the manuals already present in `docs/manuals/`, and — per mode — the `INDEX.md` structure it would overwrite or the count of manuals the dossier would generate. It does **not** run `Write`; numbering queries use `aw next-number --dry-run` (pure).
 
 ## Inputs
 
 **`agent-workflow` CLI (alias `aw`)** — never read hardcoded paths:
 
-- `aw sessions` / `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates the corpus.
+- `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates the corpus (ALL sessions, closed + active). `aw sessions` alone lists only ACTIVE sessions — never use it as the corpus.
 - `aw session-artifacts --code <NNN> --dump objetivo,decisiones` — returns `{path, content, size}` per artifact (`SESSION` with its `## Objective`, `DECISION`); the plan-doc is read by its path.
 - `aw next-number docs/manuals` — deterministic numbering (`regenerate` mode only).
 
@@ -60,7 +60,7 @@ In plan mode it **describes**, never writes: the resolved mode, the detected top
 - `docs/manuals/INDEX.md` — re-generable (overwritable) in `complement` mode.
 - The declared sources' code — read to describe behavior.
 
-**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/SKILL.md`](../../harness/SKILL.md)):
+**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/HARNESS.md`](../../harness/HARNESS.md)):
 
 ```
 /w:export-manuals [--sessions NNN[,NNN]] [--since sessionNNN] [--source <alias>]
@@ -89,7 +89,7 @@ No args: `--mode complement` over the whole corpus.
 
 ### Step 1 — Resolve context and corpus
 
-`aw sessions` / `release-data` applying `--sessions`/`--since`/`--source`. The CLI handles destination-folder resolution.
+`aw release-data` applying `--sessions`/`--since`/`--source`. The CLI handles destination-folder resolution.
 
 ### Step 2 — Inspect the present manuals
 
@@ -124,4 +124,4 @@ With `--dry-run`: print the report; write nothing. Otherwise: `complement` → `
 - Design: `docs/referencias/workflow-exports/export-manuals.md` · family: [`../README.md`](../README.md).
 - Writing: **ambient** convention (not a role) — the host auto-applies an installed writing skill when present.
 - Source artifacts: `DECISION` + plan-doc (see `docs/referencias/workflow-artifacts/artifacts-exec/` and `docs/specs`/`docs/plans`).
-- Siblings: [`../export-scripts/SKILL.md`](../export-scripts/SKILL.md) · [`../export-diagrams/SKILL.md`](../export-diagrams/SKILL.md) · [`../export-reports/SKILL.md`](../export-reports/SKILL.md).
+- Siblings: [`../export-scripts/EXPORT.md`](../export-scripts/EXPORT.md) · [`../export-diagrams/EXPORT.md`](../export-diagrams/EXPORT.md) · [`../export-reports/EXPORT.md`](../export-reports/EXPORT.md).

package/skills/w/exports/export-reports/{SKILL.md → EXPORT.md}

@@ -43,13 +43,13 @@ The report's prose follows the **ambient** writing conventions: the host auto-ap
 
 ## Read-only sandbox
 
-In plan mode it **describes**, never writes: the resolved audience/length, the corpus sessions that would enter after the filters, the sections that would appear, the R-items that would consolidate (and detected conflicts), and the estimated length. It does **not** run `Write` or effectful `aw next-number`.
+In plan mode it **describes**, never writes: the resolved audience/length, the corpus sessions that would enter after the filters, the sections that would appear, the R-items that would consolidate (and detected conflicts), and the estimated length. It does **not** run `Write`; numbering queries use `aw next-number --dry-run` (pure).
 
 ## Inputs
 
 **`agent-workflow` CLI (alias `aw`)** — never read hardcoded paths:
 
-- `aw sessions` / `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates + filters the corpus.
+- `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates + filters the corpus (ALL sessions, closed + active, with `release_eligible`). `aw sessions` alone lists only ACTIVE sessions — never use it as the corpus.
 - `aw session-artifacts --code <NNN> --dump objetivo,conclusiones,decisiones` — returns `{path, content, size}` for `SESSION` (referenced spec), `CONCLUSIONS` and `DECISION`; the plan-doc state is read by its path.
 - `aw next-number docs/reports` — deterministic numbering (the CLI handles destination-folder resolution).
 
@@ -57,7 +57,7 @@ In plan mode it **describes**, never writes: the resolved audience/length, the c
 
 - `docs/specs`, `docs/plans`, `docs/reports/*` — context + collision avoidance.
 
-**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/SKILL.md`](../../harness/SKILL.md)):
+**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/HARNESS.md`](../../harness/HARNESS.md)):
 
 ```
 /w:export-reports [--sessions NNN[,NNN]] [--since sessionNNN] [--source <alias>]
@@ -79,7 +79,7 @@ No args: the whole corpus, executive audience by default.
 
 ### Step 1 — Resolve context and filter the corpus
 
-`aw sessions` / `release-data` applying `--sessions`/`--since`/`--source`. If the resulting set is empty → **abort** with an explicit message (no sessions in the declared range). The CLI handles destination-folder resolution.
+`aw release-data` applying `--sessions`/`--since`/`--source`. If the resulting set is empty → **abort** with an explicit message (no sessions in the declared range). The CLI handles destination-folder resolution.
 
 ### Step 2 — Collect per-session inputs
 
@@ -91,7 +91,7 @@ Extract the R-items from `CONCLUSIONS`/`DECISION` (pending, deferred, "next step
 
 ### Step 4 — Synthesize (prose: ambient conventions)
 
-Render applying the ambient writing conventions (host): Executive summary · What was done (grouped by business capability, **not** by session) · Impacted components (table) · Key decisions · Results/conclusions · Pending/Roadmap (only with R-items). Technical→executive translation and length cap per `--audience`. Optional: a simple synthesis `flowchart LR` (with a `mermaid.ink` link); the detailed technical diagram belongs to `export-diagrams`.
+Render applying the ambient writing conventions (host): Executive summary · What was done (grouped by business capability, **not** by session) · Impacted components (table) · Key decisions · Results/conclusions · Pending/Roadmap (only with R-items). Technical→executive translation and length cap per `--audience`. Optional: a simple synthesis `flowchart LR` (inline; a `mermaid.ink` link is OPTIONAL — it encodes the diagram source into a public-service URL, omit it for private corpora); the detailed technical diagram belongs to `export-diagrams`.
 
 ### Step 5 — Write or report
 
@@ -110,4 +110,4 @@ Functionally idempotent: each invocation takes the next `NNN`; it never overwrit
 - Design: `docs/referencias/workflow-exports/export-reports.md` · family: [`../README.md`](../README.md).
 - Writing: **ambient** convention (not a role) — the host auto-applies an installed writing skill when present.
 - Inputs: spec (`docs/specs`), `CONCLUSIONS`/`DECISION` (see `docs/referencias/workflow-artifacts/`), plan-doc (`docs/plans`).
-- Siblings: [`../export-scripts/SKILL.md`](../export-scripts/SKILL.md) · [`../export-manuals/SKILL.md`](../export-manuals/SKILL.md) · [`../export-diagrams/SKILL.md`](../export-diagrams/SKILL.md).
+- Siblings: [`../export-scripts/EXPORT.md`](../export-scripts/EXPORT.md) · [`../export-manuals/EXPORT.md`](../export-manuals/EXPORT.md) · [`../export-diagrams/EXPORT.md`](../export-diagrams/EXPORT.md).

package/skills/w/exports/export-scripts/{SKILL.md → EXPORT.md}

@@ -44,21 +44,22 @@ The **`sql`** capability (built-in default `sql`), resolved via `.workflow/skill
 
 ## Read-only sandbox
 
-In plan mode it **describes**, never writes: the resolved `NNN`, the detected sources (sessions + standalone), the categories with content, the files that would appear at the bundle root and the approximate README content. It does **not** run `Write`, effectful `aw next-number`, or mutations.
+In plan mode it **describes**, never writes: the resolved `NNN`, the detected sources (sessions + standalone), the categories with content, the files that would appear at the bundle root and the approximate README content. It does **not** run `Write` or mutations; numbering queries use `aw next-number --dry-run` (pure).
 
 ## Inputs
 
 **`agent-workflow` CLI (alias `aw`)** — never read hardcoded paths:
 
-- `aw sessions` / `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates the session corpus.
+- `aw release-data [--since sessionNNN] [--source <alias>]` — enumerates the session corpus (ALL sessions, closed + active, with `release_eligible`). `aw sessions` alone lists only ACTIVE sessions — never use it as the corpus.
 - `aw session-artifacts --code <NNN> --dump scripts` — lists the session's `.sql` files with path and size (content is read by path). No scripts → empty list, silent skip.
-- `aw next-number docs/scripts` — deterministic numbering of the bundle directory (the CLI handles destination-folder resolution).
+- `aw release-data --standalone-sql` — lists the loose top-level `docs/scripts/*.sql` (source B) deterministically, with `is_rollback` flag; `--include-graduated` lists previous bundles (modern `NNN-export-scripts-YYYY-MM-DD` and legacy naming) for the exclusion/dedup step.
+- `aw next-number docs/scripts` — deterministic numbering of the bundle directory; it also creates `docs/scripts` when missing (the CLI guarantees destination resolution). In plan mode use `--dry-run` (pure query).
 
 **Filesystem**:
 
-- Standalone `docs/scripts/*.sql` (top-level only), **excluding** any `docs/scripts/NNN-export-scripts-*/` (previous outputs of this export).
+- Standalone `docs/scripts/*.sql` (top-level only, via `--standalone-sql`), **excluding** any `docs/scripts/NNN-export-scripts-*/` (previous outputs of this export).
 
-**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/SKILL.md`](../../harness/SKILL.md)):
+**Args** (no lifecycle *structured-choice*; harness capability — see [`../../harness/HARNESS.md`](../../harness/HARNESS.md)):
 
 ```
 /w:export-scripts [--sessions NNN[,NNN]] [--since sessionNNN] [--source <alias>]
@@ -79,7 +80,7 @@ No args: every corpus session + every standalone `.sql` (excluding previous bund
 
 ### Step 1 — Collect SQL sources
 
-**Source A — sessions**: for every corpus session (`aw sessions` / `release-data` + `session-artifacts --code <NNN> --dump scripts`), read the `.sql` files the dump lists (per-script path). Take **only** type-B statements (deliverable DDL/DML migrations); ignore read-only type-A (diagnostic queries). Expected per-statement markers: `-- @category: <01-04>` + `-- @stmt: NNN-verb-target` (format defined by the `sql` capability).
+**Source A — sessions**: for every corpus session (`aw release-data` + `session-artifacts --code <NNN> --dump scripts`), read the `.sql` files the dump lists (per-script path). Take **only** type-B statements (deliverable DDL/DML migrations); ignore read-only type-A (diagnostic queries). Expected per-statement markers: `-- @category: <01-04>` + `-- @stmt: NNN-verb-target` (format defined by the `sql` capability).
 
 **Source B — standalone** (unless `--skip-standalone`): list top-level `docs/scripts/*.sql`, **excluding** `docs/scripts/NNN-export-scripts-*/`. Per file: honor `@category` markers when present; otherwise infer the category from content (`CREATE/ALTER TABLE`, `CREATE INDEX` → `01`; `CREATE OR REPLACE FUNCTION`/`PROCEDURE` → `02`; `UPDATE`/`DELETE` → `03`; `INSERT INTO … VALUES` → `04`). If the filename contains `rollback` → skip (it never enters a forward).
 
@@ -132,4 +133,4 @@ Functionally idempotent: each invocation takes the next `NNN` and **never overwr
 - Design: `docs/referencias/workflow-exports/export-scripts.md` · family: [`../README.md`](../README.md).
 - Composed capability: `sql` (built-in default; see `docs/referencias/workflow-roles/`).
 - Source artifact: `SCRIPTS.sql` (see `docs/referencias/workflow-artifacts/artifacts-core/`).
-- Siblings: [`../export-manuals/SKILL.md`](../export-manuals/SKILL.md) · [`../export-diagrams/SKILL.md`](../export-diagrams/SKILL.md) · [`../export-reports/SKILL.md`](../export-reports/SKILL.md).
+- Siblings: [`../export-manuals/EXPORT.md`](../export-manuals/EXPORT.md) · [`../export-diagrams/EXPORT.md`](../export-diagrams/EXPORT.md) · [`../export-reports/EXPORT.md`](../export-reports/EXPORT.md).

package/skills/w/harness/{SKILL.md → HARNESS.md}

@@ -49,7 +49,7 @@ Concrete mechanism per harness (**Jul-2026**, verified against official docs; `~
 
 | Capability | Claude Code | Codex | Gemini / Antigravity | OpenCode | Crush | Warp / Oz | Generic |
 |---|---|---|---|---|---|---|---|
-| command-invocation | `.claude/commands/` (slash) | slash + skills | `.gemini/commands/*.toml` | `.opencode/command/` | user-invocable skills | Workflows (Drive) | text |
+| command-invocation | `.claude/commands/` (slash) | skills only (`$` mention; no commands dir, prompts removed) | skills only in agy (system slash commands; `.gemini/commands/*.toml` = legacy Gemini CLI) | `.opencode/command/` | `.crush/commands` (palette) + user-invocable skills | skills as `/name` | text |
 | procedure-loading (skills) | `SKILL.md` `.claude/skills` | `SKILL.md` `.agents/skills` | `SKILL.md` (agentskills) | `SKILL.md` `.opencode`+`.claude`+`.agents` | `SKILL.md` `.agents`+`.crush`+`.claude` | `SKILL.md` `.agents`+`.warp`+`.claude` | read-and-follow `.md` |
 | structured-choice | `AskUserQuestion` (**main-agent only**) | — | — | — | — | — | numbered markdown |
 | compaction | `/compact` | Pre/PostCompact hooks | ~ | `session.compacted` | ~ | ~ | CHECKPOINT + resume |
@@ -61,7 +61,7 @@ Concrete mechanism per harness (**Jul-2026**, verified against official docs; `~
 
 > **Notes (field research Jul-2026):** **`SKILL.md` skills** are the **universal** portable unit — **all six** harnesses support them (Codex added them Dec-2025; **`.agents/skills` is the cross-host anchor**, read by Codex/OpenCode/Crush/Warp). **Structured choice** (`AskUserQuestion`) remains **Claude Code / main-agent only** → elsewhere `structured-choice` degrades to numbered markdown. The **enforcement layer** (new row) is **NO longer Claude-exclusive**: Codex + Gemini use a near-identical protocol (`permissionDecision:deny` / exit 2) and OpenCode blocks via `throw` in a JS plugin; Crush/Warp only offer **coarse** allow/deny (no custom per-command logic) → there, conventions stay **advisory** + allow/deny lists. Enforced **plan mode** is never trusted for safety; git-safe (invariant #5) is our own. **MCP** is universal (each host its file/key). The **guaranteed floor** (last column) runs the full model.
 
-> **Oz (Warp's cloud sibling).** `oz agent run` is a cloud agent orchestrator that **reuses Warp's surfaces**: same skills (`.agents/skills`, flattened to top level like Warp) and `AGENTS.md`, with `structured-choice` equally degraded to numbered markdown. It differs in three points: **detection** via `OZ_RUN_ID` (takes priority over Warp when both markers coexist); **MCP without a config file** — the JSON is passed via the `--mcp` flag of `oz agent run` (or the `OZ_MCP_CONFIG` env), it never writes `.warp/.mcp.json`; and **no plugin or hooks** (advisory enforcement, like Warp). Hence it shares the **Warp / Oz** column with that MCP caveat.
+> **Oz (Warp's cloud sibling).** `oz agent run` is a cloud agent orchestrator that **reuses Warp's surfaces**: same skills (`.agents/skills`, top-level dirs like Warp) and `AGENTS.md`, with `structured-choice` equally degraded to numbered markdown. It differs in three points: **detection** via `OZ_RUN_ID` (takes priority over Warp when both markers coexist); **MCP without a config file** — the JSON is passed via the `--mcp` flag of `oz agent run` (or the `OZ_MCP_CONFIG` env), it never writes `.warp/.mcp.json`; and **no plugin or hooks** (advisory enforcement, like Warp). Hence it shares the **Warp / Oz** column with that MCP caveat.
 
 ## Leverage installed skills
 
@@ -72,7 +72,7 @@ Concrete mechanism per harness (**Jul-2026**, verified against official docs; `~
 
 ## Convention for the rest of the corpus
 
-- Loops/commands reference the **capability** by name (e.g. "*structured-choice* (see `harness/SKILL.md`)"), **never** the concrete tool.
+- Loops/commands reference the **capability** by name (e.g. "*structured-choice* (see `harness/HARNESS.md`)"), **never** the concrete tool.
 - The historical name `AskUserQuestion` survives **only** as the Claude-Code binding of `structured-choice` (this table), never as doctrine vocabulary.
 - The `flow` lifecycle control (`Compactar`/`Cerrar`) belongs to the `structured-choice` capability, not to a tool: on harnesses without structured choice it is offered as one more textual option.
 
@@ -82,7 +82,18 @@ Proven pattern (Spec Kit, 30+ agents): **one canonical source** + generate/symli
 
 ## Command packaging (harness-specific)
 
-Each command's **contract** (Flow, Trigger, Input, Mode, …) is agnostic. The **file** the harness executes wraps that contract in its native format: Claude Code = slash command with frontmatter (`description`, `argument-hint`, `allowed-tools`) + a body that invokes the skill or the `aw` CLI; Codex = slash command or skill; Gemini/Antigravity = `.gemini/commands/*.toml`; OpenCode = `.opencode/command/*.md`; Crush/Warp = the invocable **skill itself**. The **universal fallback** is *skill-as-command*: since all six harnesses load skills, the doctrine always runs even without native commands. The contract never changes; the wrapper does (another column).
+Each command's **contract** (Flow, Trigger, Input, Mode, …) is agnostic. The **file** the harness executes wraps that contract in its native format — the installer (`aw self install-skill`) emits the right wrapper per host:
+
+| Host | Wrapper installed | Invoked as |
+|---|---|---|
+| Claude Code | `~/.claude/commands/w/<cmd>.md` (frontmatter `description`/`argument-hint`/`allowed-tools`) | `/w:<cmd>` |
+| Codex | synthesized skill `~/.codex/skills/w-<cmd>/SKILL.md` (Codex reads no commands dir; custom prompts deprecated/removed since 0.14x) | `$w-<cmd>` mention |
+| Gemini/Antigravity | synthesized skill `~/.gemini/skills/w-<cmd>/SKILL.md` (agy reads NO commands dir — slash commands are system-only; verified vs agy 1.0.16 binary) + `~/.gemini/commands/w/<cmd>.toml` kept for legacy Gemini CLI | skill (agy) · `/w:<cmd>` (legacy CLI) |
+| OpenCode | `~/.opencode/command/w/<cmd>.md` | `/w/<cmd>` |
+| Crush | `~/.crush/commands/w/<cmd>.md` (plain body — Crush parses no frontmatter) | palette `user:w:<cmd>` |
+| Warp/Oz | synthesized skill `w-<cmd>/SKILL.md` next to the bundle (Warp lists skills as `/name`) | `/w-<cmd>` |
+
+*Skill-as-command* (a synthesized `w-<cmd>` skill whose body is the command, with bundle references rewritten to `../w/…`) is the **universal fallback** for any host without a native commands surface. The loop/role/export manuals are deliberately **not** `SKILL.md` files (`LOOP.md`/`ROLE.md`/`EXPORT.md`/`HARNESS.md`): hosts that scan skill roots **recursively** (Codex ≤6 levels; OpenCode and Crush — which also cross-read `~/.claude/skills` and `~/.agents/skills`) must never index the internals as invocable skills. The contract never changes; the wrapper does (another column).
 
 ## Status
 

package/skills/w/loops/CHASSIS.md

@@ -6,11 +6,11 @@ This document is the **common engine** of the agent-workflow loops: the doctrine
 
 The **5 loops** run this engine; each adds only its deltas:
 
-- [`spec-refine-loop`](spec-refine-loop/SKILL.md) — refines the **spec** in place; deltas: spec gap taxonomy, analyze gate, `## UI spec` via the `ui-design` capability.
-- [`plan-new-loop`](plan-new-loop/SKILL.md) — generates the **plan** from the spec; deltas: rich plan + plan gap taxonomy (+ per-screen design SPECs when the plan includes UI).
-- [`plan-refine-loop`](plan-refine-loop/SKILL.md) — refines the **plan** in place (auxiliary, not mandatory); reuses the gap taxonomy + coherence gate of `plan-new-loop`. It is to `plan-new` what `spec-refine` is to `spec-new`.
-- [`plan-exec-loop`](plan-exec-loop/SKILL.md) — **executes** the plan: code/DB/git, a single session per run, per-phase progress in the plan-doc, no auto-export. Applies the policies in [`CODE-POLICIES.md`](CODE-POLICIES.md).
-- [`quick-loop`](quick-loop/SKILL.md) — the engine with **minimal ceremony** (the prompt *is* the objective); also applies [`CODE-POLICIES.md`](CODE-POLICIES.md) (proportional gate).
+- [`spec-refine-loop`](spec-refine-loop/LOOP.md) — refines the **spec** in place; deltas: spec gap taxonomy, analyze gate, `## UI spec` via the `ui-design` capability.
+- [`plan-new-loop`](plan-new-loop/LOOP.md) — generates the **plan** from the spec; deltas: rich plan + plan gap taxonomy (+ per-screen design SPECs when the plan includes UI).
+- [`plan-refine-loop`](plan-refine-loop/LOOP.md) — refines the **plan** in place (auxiliary, not mandatory); reuses the gap taxonomy + coherence gate of `plan-new-loop`. It is to `plan-new` what `spec-refine` is to `spec-new`.
+- [`plan-exec-loop`](plan-exec-loop/LOOP.md) — **executes** the plan: code/DB/git, a single session per run, per-phase progress in the plan-doc, no auto-export. Applies the policies in [`CODE-POLICIES.md`](CODE-POLICIES.md).
+- [`quick-loop`](quick-loop/LOOP.md) — the engine with **minimal ceremony** (the prompt *is* the objective); also applies [`CODE-POLICIES.md`](CODE-POLICIES.md) (proportional gate).
 
 ## Persistent objective
 
@@ -92,7 +92,7 @@ The **CLI owns the number**: `aw session-create` prepends a **global, sequential
 
 - `aw session-create --type <type> --name <slug>-<flow>` → creates `NNN-<slug>-<flow>` / `aw session-resume --code <…>` (detects `CHECKPOINT`).
 - `aw checkpoint-write` / `aw checkpoint-read` for resume.
-- `aw session-close` on close (with reason); `aw session-artifacts` to inspect.
+- `aw session-close` on close — also upserts the session's HISTORY.md row (the durable record; sessions/ is gitignored). Non-fatal: on `history_error` in its output, re-run `aw history-update --code <NNN> --state closed`. `aw session-artifacts` to inspect.
 - **Reopen to continue** (operating context, row 2): `aw session-resume --code <NNN> --reopen` reactivates a **closed** session (removes `.closed` → active) to keep working in it; without `--reopen`, resume is read-only. To detect the most recent closed one: `aw resume-summary --include-recent-closed` (or `aw sessions --state all`).
 
 ## Ask-vs-research rule (the discriminator)
@@ -119,7 +119,7 @@ Investigation is **inline**: an activity **inside the run's current session**, n
 
 ## Structured-choice (design & batching)
 
-**Canonical rule (single source — the rest of the corpus only references it):** *structured-choice* = **≤3 content questions + 1 `flow` control**, always. Per-harness binding in [`../harness/SKILL.md`](../harness/SKILL.md) (Claude Code: `AskUserQuestion`, max 4 questions/call; without structured choice it degrades to **numbered markdown**).
+**Canonical rule (single source — the rest of the corpus only references it):** *structured-choice* = **≤3 content questions + 1 `flow` control**, always. Per-harness binding in [`../harness/HARNESS.md`](../harness/HARNESS.md) (Claude Code: `AskUserQuestion`, max 4 questions/call; without structured choice it degrades to **numbered markdown**).
 
 - Since the `flow` control is **always** present → **≤3 content questions + 1 `flow` control**.
 - **`flow` control** (lifecycle, always present): `Compactar` | `Cerrar`. Answering only the content questions (not touching `flow`) = keep iterating.
@@ -142,7 +142,7 @@ Resume **keys off the `CHECKPOINT`** of the run's session, not the existence of
 
 > Each heir defines its **prior-work mark**: in the refine loops, the presence of `## Refinement decisions` + `## Q&A traceability` in the doc; in plan-exec, the plan-doc's `- [x]` checkboxes; quick has no doc (resume by CHECKPOINT only).
 
-> **`Compactar`** (the `flow` control, across all 3 cases) → write `CHECKPOINT.md` in the session (in-flight progress, remaining gaps, Q&A, `attempts`) → trigger the harness **compaction** (Claude Code: `/compact`; see [`../harness/SKILL.md`](../harness/SKILL.md)) → resume by reading the checkpoint.
+> **`Compactar`** (the `flow` control, across all 3 cases) → write `CHECKPOINT.md` in the session (in-flight progress, remaining gaps, Q&A, `attempts`) → trigger the harness **compaction** (Claude Code: `/compact`; see [`../harness/HARNESS.md`](../harness/HARNESS.md)) → resume by reading the checkpoint.
 
 ## Convergence / exit
 
@@ -162,7 +162,7 @@ The loops that **edit code** (`plan-exec-loop`, `quick-loop`) additionally run t
 Applies to **every** relative reference in the doctrine — never repeated per link:
 
 1. **Normal install** (the `w/` tree): the relative path resolves as-is (`../CHASSIS.md`, `../../commands/spec-new.md`).
-2. **Flattened install** (e.g. Warp/Oz): the shared `.md` files (`CHASSIS.md`, `CODE-POLICIES.md`) sit **next to the loop's `SKILL.md`**; another loop is a **sibling** skill `w-<loop>/` (e.g. `../spec-refine-loop/SKILL.md` → `../w-spec-refine-loop/SKILL.md`).
+2. **Synthesized command skills** (hosts without a commands dir — Codex/Warp/Oz): each command installs as a **sibling** skill `w-<command>/` with its references rewritten into the bundle (`../loops/…` → `../w/loops/…`); the `w/` tree stays intact, so loop-to-loop references resolve as a normal install.
 3. A reference that does not resolve = **optional deep-dive** — this engine's doctrine is self-contained.
 
 The chassis **is not a skill** (no frontmatter; never invoked nor bound via `.workflow/skills.toml`): it enters the context only because a loop orders it read from its `## Inherits`. It does not define flow, deliverable or gap taxonomy — that belongs to each heir.