---
name: agami-query
description: "Answers natural-language questions about the user's database. Loads the agami semantic model (subject areas, tables, columns, relationships with join cardinality, entities, metrics) and few-shot examples from <artifacts_dir>/<profile>/, generates SQL via the examples-first traversal (pick subject area → match examples → resolve entities/metrics → compound table context), executes it locally via the user's chosen tool (psql / mysql / snowsql / sqlite3 native CLI, DuckDB binary, or the Python driver `execute_sql.py` — which runs the scope gates and reports fan-trap/chasm-trap and aggregation findings on the receipt), returns results as a markdown table with optional CSV export, and renders Chart.js HTML charts on request. All execution is local — no data leaves the machine."
when_to_use: "Use when the user asks 'how many', 'show me', 'top N', 'trend over time', 'compare', 'breakdown by', 'group by', 'average', or any other data question against their configured database. Also use for CSV export ('export this'), chart rendering ('make that a bar chart'), or to follow up on a previous result ('drill into the EU region')."
argument-hint: "[question] [--csv] [--chart bar|line|pie|doughnut|scatter]"
---

# agami query-database

You answer the user's natural-language question about their database. Goal: generate correct SQL from the semantic model + the few-shot examples via the examples-first traversal, execute it locally, return rows + an insight, and offer a chart / export when appropriate. Everything runs on the user's machine.

This skill orchestrates:

1. **Setup** (once per session) — resolve the profile + the semantic model at `<artifacts_dir>/<profile>/`, verify the configured database tool still works.
2. **Generate SQL** — examples-first traversal: pick the subject area → match curated examples → (cold start) resolve entities/metrics + identify opaque literals → compound `get_table_context` → produce one SQL statement → safety checks.
3. **Execute** — run via the chosen tool; the Python tier runs the fan/chasm pre-flight and the scope/PII gates; auto-retry on classified errors; risk-assess large-table queries.
4. **Present** — markdown table; CSV via `--csv` or "export this"; Chart.js HTML via `--chart` or "make that a chart".
5. **Log + post-install GitHub-star ask** — write `<artifacts_dir>/local/query_log.jsonl` and ask the user (once, after first successful query) to star us on GitHub; once they answer, point them to `/agami-serve` (wire the model into Claude Desktop — the experience their business users get).

For the model format: [`semantic_model/__init__.py`](../../../../packages/agami-core/src/semantic_model/__init__.py) (layout) + `packages/agami-core/src/semantic_model/models.py`.
For SQL safety: [`shared/sql-generation-rules.md`](../../shared/sql-generation-rules.md).
For dialect-specific syntax: [`shared/dialect-rules.md`](../../shared/dialect-rules.md).
For connection method + execution: [`shared/connection-reference.md`](../../shared/connection-reference.md).
For DB error classification: [`shared/db_error_classifier.md`](../../shared/db_error_classifier.md).
For chart template: [`shared/chart-template.html`](../../shared/chart-template.html).

## Invocation conventions

**Read [`shared/invocation-conventions.md`](../../shared/invocation-conventions.md) before suggesting any slash command in chat.** Agami slash commands: `/agami-connect`, `/agami-query`, `/agami-model`, `/agami-save-correction`, `/agami-reconcile`. (`/agami-model`'s Review tab absorbed the former `/agami-review`.) Never write the un-prefixed forms (`/init`, `/connect`, `/query-database`, etc.) or colon-namespaced forms (`/agami:init`, etc.) — those don't exist. **`/agami-init` was folded into `/agami-connect` Phase 0a** — credential setup now lives there.

For chat replies, **prefer natural language over slash commands** — it reads better and the skill's `when_to_use` matcher routes correctly:

- Re-introspect the schema → "say 'reload the schema'" or "say 'reintrospect my database'"
- Save a correction → "say 'save this as a correction'" or "say 'remember this'"
- Ask a data question → just type the question
- Set up agami / switch profiles → `/agami-connect` (the one place the slash form is genuinely cleaner than natural language — agami-connect handles credentials too via Phase 0a)

## Conversation style

- **One question per turn unless they're truly bundled.**
- **Use AskUserQuestion sparingly** — only when the user must pick before the skill can proceed (large-table HIGH-risk approval, the post-install GitHub-star ask, the demo-query Yes/No/Skip in agami-connect). **Do NOT use AskUserQuestion for follow-up suggestions** — those are 5 plain numbered bullets per Phase 4f.
- **Insights, not narration** — lead with the answer ("Carol Chen has the highest spend at $148.95"), not the SQL or the process.
- **Round numbers in prose**, exact in the table.
- **Don't echo the SQL in chat prose** — that's enforced as a hard rule in Phase 2. Don't paste the raw Bash CSV — Phase 3.

---

## Phase −1: Plan-mode check

Run the detection + ask logic from [`shared/plan-mode-check.md`](../../shared/plan-mode-check.md). agami-query needs Bash (SQL execution) and Write (chart HTML) — both are blocked in plan mode.

**If plan mode is active and the user picks `Stay in plan mode`:**

- **Reopen-last-chart intent** (Phase 2a.1 below) — re-displaying an existing HTML chart only needs `Read` plus `open <path>`. Run that flow if matched.
- **Anything else** — refuse and end the turn. **DO NOT write a plan file. DO NOT call `ExitPlanMode`.** Refusal text (verbatim):

  > I can't run SQL in plan mode. Switch to **Auto** or **Edit Automatically** mode (Shift+Tab to cycle) and re-invoke me.

If plan mode is not active, skip this phase silently and go to Phase 1.

---

## Phase 1: Setup (once per session)

### HARD RULES — connection rules

These are non-negotiable.

1. **Connect ONLY to the host/port/database/user/password in `<artifacts_dir>/local/credentials`** — the sole credential source (no env-var bypass). Never substitute `localhost` or any other host as a fallback. Never connect to anywhere not in the credentials.
2. **Never ask the user for connection details in chat.** If credentials are missing, stop and invoke `/agami-connect` — its Phase 0a runs the DB-type picker, writes `<artifacts_dir>/local/credentials.example`, and ends the turn for the user to fill it in.
3. **Never scan or guess.** No `pgrep`, no `ps`, no `find /`, no `ls /Applications/Postgres.app`, no listing port-listeners. The only Bash probes allowed during setup are `which <tool>` for a database tool on `PATH` and `python3 -c 'import <module>'` for a Python driver.
4. **NEVER put the password (or any credential field) in a Bash command line.** That includes `export PGPASSWORD='<value>'`, `export MYSQL_PWD='<value>'`, `psql -W <password>`, `mysql -p<password>`, or any heredoc / stdin form that interpolates the password. Hosts render Bash tool calls as collapsibles in their UI — anything in the command becomes visible in the chat. Use the auth files generated by `scripts/setup_pgauth.py` (see [`shared/connection-reference.md → HARD RULES`](../../shared/connection-reference.md)). For native CLI queries the visible Bash command is `PGPASSFILE=<artifacts_dir>/local/.pgpass psql -h ... -U ... -d ... -c "$SQL" --csv`. For the Python driver path use `"$PY" -m execute_sql --sql-file ...`.

These rules apply to every phase of this skill, not just Phase 1.

### 1a — credentials check (binding)

Read `<artifacts_dir>/local/credentials`. If the file (or the active profile's section) is missing, invoke `/agami-connect` (its Phase 0a handles first-time credential setup) and **stop this skill**. Do not continue to load the semantic model. Do not run any other Bash commands.

### 1b — load the semantic model

Resolve `<profile>`: `AGAMI_PROFILE` → `active_profile` in `<artifacts_dir>/local/.config` → `"main"`.
Resolve `<artifacts_dir>` per [`shared/file-layout.md`](../../shared/file-layout.md#configuring-artifacts_dir): `AGAMI_ARTIFACTS_DIR` → `.config.artifacts_dir` → `$HOME/agami-artifacts`.

The model is the semantic-model tree at `<artifacts_dir>/<profile>/` (`datasource.yaml` + `subject_areas/<area>/…`). There is no legacy-layout fallback — the model is the only format.

**Never hand-read OR hand-roll the model.** Don't `cat`/`Read` `datasource.yaml`, `subject_areas/**`, `tables/*.yaml`, or `relationships.yaml`, and **never write a `python -c` / ad-hoc script to load, dump, or walk the model tree** — that guesses the schema, breaks on a wrong key, and can leak a traceback to the user. The CLI returns the same data structured, and the layout is already known (relationships + entities + metrics live at the **area** level, not inside a table file). `sm areas "$ROOT"` is the **one-call model map**: per area it returns `table_count`, `entity_count`, `metric_count`, `relationship_count` + description — the whole shape in a single call. (Column-level detail → `sm context`; browsable table/column tree → `sm model-tree`.) **When the user asks "what does the model look like" / "show me the model," run `sm model-tree` (or open `/agami-model`) — don't improvise Python.**

**Don't run a separate existence probe either** (no `ls datasource.yaml`, and never probe for the plugin's own scripts — `sm`, `execute_sql.py`, `semantic_model/` always ship with the plugin). That same first `sm areas` call doubles as the check: model present → you get the map; absent → the CLI returns `{"error":"no_model"}` with **exit code 3** → invoke `agami-connect` and stop.

Drive everything through the CLI — the `sm` wrapper resolves the interpreter + deps. (These granular steps are CLI operations; on the MCP surface they're folded into the smart `get_datasource_schema`, which advertises the 4 product tools — so don't invoke the steps below as MCP tools.)

```bash
ROOT="<artifacts_dir>/<profile>"
bash "$AGAMI_PLUGIN_ROOT/scripts/sm" areas "$ROOT"; rc=$?      # subject-area index; rc 3 = no model → agami-connect
bash "$AGAMI_PLUGIN_ROOT/scripts/sm" context "$ROOT" --area A --tables t1 t2   # compound table context
bash "$AGAMI_PLUGIN_ROOT/scripts/sm" examples "$ROOT" --area A --query "…"     # examples-first ranking
```

The model loader already drops `review_state: rejected` entries from what it serves and applies the area's `expose_column_groups` scoping, so you never see excluded tables/columns/relationships. (Rejections are the curator's choice via `/agami-model` — surfaced nowhere.) When a query would touch a `stale` entry, warn once: *"This would use `<entity>`, marked stale (schema drift). Run /agami-connect to re-introspect, then /agami-model to reconcile."*

### 1c — what the model gives you

You don't build hand-rolled indexes — the loader returns structured objects. The pieces you'll use during SQL generation:

- **Subject areas** — the primary scoping unit (replaces "load every table"). Each has a description, a table list, entities, metrics, and an intra-area relationship graph (each edge carries join **cardinality** + a trust block).
- **`get_table_context(area, tables)`** — columns (scoped by `expose_column_groups`), `default_filters`, relationships, `caveats`, `value_transforms`, metrics — in one call.
- **Entities** — the vocabulary users say (name/plural/other_names → `maps_to` table.column, with a `value_pattern` for opaque IDs). Use `resolve_entities` / `identify_entity`.
- **Metrics** — reusable aggregations with prose `calculation` + per-dialect `bindings`. **Use the binding SQL VERBATIM** when the user asks for a metric by name or synonym; don't hand-roll the aggregate.
- **Cross-subject-area relationships** (org level) — for joins that span two areas.

### 1d — load the examples library

Examples live per subject area at `<artifacts_dir>/<profile>/prompt_examples/<area>/examples.yaml`. Use `cli examples "$ROOT" --area <area> --query "<question>"` to rank them (the **examples-first** signal — step 2a). If a high-confidence match returns, mirror its tagged tables/columns/SQL shape and skip cold-start resolution.

If there are no examples for the relevant area → warn: "I don't have few-shot examples for this database yet — answers may be lower quality. Say 'introspect the schema' to seed them." (Slash form `/agami-connect` only if the user asks "what do I type?".)

### 1d.1 — load USER_MEMORY.md

Read `<artifacts_dir>/USER_MEMORY.md` (if present). Strip HTML comments (`<!--...-->`), then keep the rest. If the file is missing, treat it as empty — never error. See [`shared/user-memory-format.md`](../../shared/user-memory-format.md) for what's in it.

This file holds free-form **user preferences across every database** (default filters, display preferences). Inject it into the SQL-generation prompt in Phase 2b under a labeled `## User memory (preferences and policies)` section — the LLM uses it as steering context.

### 1d.2 — load domain context

Run `cli org-context "$ROOT"` — it returns the **full** domain context for this database in one block: the human's datasource.md narrative (HTML comments stripped) **plus** the model-derived summary that the file does NOT contain — subject areas, conventions, and the **decoded glossary** (`key_terminology` + enum legends), assembled fresh from the structured model. Don't `Read` datasource.md by hand: the file holds only the human narrative; the glossary and summary live in the model, and this command is the one that combines them. If there's no model, treat as empty — never error. See [`shared/organization-context-format.md`](../../shared/organization-context-format.md).

Inject the result into the SQL-generation prompt in Phase 2b under `## Datasource context`, **before** the `## User memory` section — domain knowledge precedes display preferences in the LLM's reading order.

Order in Phase 2b prompt:
1. Schema context (tables / columns / relationships / metrics from the semantic model)
2. `## Datasource context` ← from `cli org-context` (narrative + derived summary + glossary)
3. `## User memory (preferences and policies)` ← from USER_MEMORY.md
4. Few-shot examples
5. The user's question

### 1e — the connection invocation pattern (do NOT run a standalone probe)

Look up the cached connection method from `<artifacts_dir>/local/.config`. **Do NOT run a separate `SELECT 1` connectivity probe** — it's a wasted round-trip (and pointless for a local SQLite/DuckDB file). The user's *actual* query is the connectivity check: run it directly, and if it fails, classify the error via [`db_error_classifier.md`](../../shared/db_error_classifier.md). The table below is the **exact invocation pattern per tier** for running that query — **don't guess flags** (`execute_sql.py` does NOT accept positional SQL, a `--format` flag, or any flag not listed; guessing produces "unrecognized arguments" errors that waste turns). The `SELECT 1` in each row is only a placeholder for *your* SQL.

| tier | invocation pattern (substitute your SQL for `SELECT 1`) |
|---|---|
| `cli` (postgres) | `PGPASSFILE="<artifacts_dir>/local/.pgpass" psql -h <host> -U <user> -d <db> -c 'SELECT 1' --csv` |
| `cli` (mysql) | `mysql --defaults-file="<artifacts_dir>/local/.mysql.cnf" --defaults-group-suffix="_<profile>" -e 'SELECT 1' --batch` |
| `cli` (snowflake) | `snowsql --config "<artifacts_dir>/local/.snowsql.cnf" -c "<profile>" -q 'SELECT 1' -o output_format=csv -o friendly=false` |
| `cli` (sqlite) | `sqlite3 -header -csv "<path>" 'SELECT 1'` — **always `-header`**, or result CSVs lose column names and `format-table` treats the first data row as the header (a wasteful re-export). |
| `duckdb` (any) | `duckdb -init "$init_file" -c 'SELECT 1' --csv` (see `build_duckdb_attach.py` for `$init_file`) |
| `python` (all DBs) | `AGAMI_PROFILE="<profile>" "$PY" -m execute_sql --sql 'SELECT 1'` |

The Python tier's CLI is **`--sql <string>`** or **`--sql-file <path>`** — those are the only two ways to pass SQL. Optional flag: `--profile <profile>` (overrides `AGAMI_PROFILE` env). **Output is RFC-4180 CSV on stdout**, always — no `--format` flag exists. If you need JSON, post-process the CSV.

Route any error through [`shared/db_error_classifier.md`](../../shared/db_error_classifier.md). Common cases:

- `auth` / `dsn` → credentials may have rotated; point at `<artifacts_dir>/local/credentials`.
- `network` → check VPN / DB endpoint reachability.
- `sign_in_required` → the person's own sign-in expired; relay the message (sign in again, new conversation).
- `driver_missing` → fall through to the next available method.

If the cached method doesn't work, re-run tool detection per [`agami-connect/SKILL.md → Phase 0a.5`](../agami-connect/SKILL.md#0a5--tool-detection).

---

## Phase 2: Generate SQL

### HARD RULE — never echo SQL in chat prose

The generated SQL belongs in two places only: (1) the Bash invocation that executes it (which the host shows as a collapsible tool call — outside our control), and (2) the collapsible "SQL" section of the HTML report written in Phase 4. **Never paste, quote, or summarize the SQL in the assistant's narrated text.** No `SELECT ...` lines, no fenced ```sql blocks, no "I'm running this query: ..." prose. Users get the SQL by clicking the SQL details element in the HTML report.

This rule applies to every retry, every fallback, every regenerate. The chat prose stays focused on approach, fetching, and insight.

### 2a — classify the input

Check intents in this order. The first match wins; only that branch runs.

1. **Reopen-last-chart intent** (handled in 2a.1 below). Triggered by short messages that ask to re-display the most recent chart without re-running SQL. Trigger phrases:
   - "reopen", "reopen the chart", "reopen that"
   - "open the last chart", "open that again", "open my last report"
   - "show me that chart again", "show me the last chart", "show that"
   - "open the previous chart", "show that report"
   - Any message ≤ 8 words that combines an open-verb (open / show / see / view / display) with a chart-noun (chart / report / it / that / last / again).

   If matched → jump to **2a.1** and skip Phases 2b–4.

2. **A question** (contains `?` or starts with how/what/show/list/which/count/give/get/find/total/average/top/which AND isn't matched by the reopen intent above) → save it as the user's data question. Continue to 2b.

3. **Empty** → ask the user; suggest 2-3 questions from the model's `ai_context.examples` if present, or inferred from `datasets[].description`.

4. **Flag-only** (`--csv` / `--chart bar`) → re-run the previous query with the flag applied.

5. **Follow-up like "make that a chart"** → see Phase 4e.

### 2a.1 — Reopen-last-chart flow (no new SQL)

If the user's intent is to re-display the most recent chart:

1. Read `<artifacts_dir>/local/query_log.jsonl` (each line is a JSON object) and take the last non-empty line **whose `status` is `"ok"` or absent** — the log also records refused and failed executions, and neither of those rendered a chart. Entries written before the log carried a `status` field have none; treat those as `ok`.
2. Look at the `chart_path` field. Possible cases:
   - **`chart_path` set AND the file exists on disk** → run `open <path>` (macOS), `xdg-open <path>` (Linux), or `start <path>` (Windows). Surface a one-liner in chat:
     ```
     Reopened: <artifacts_dir>/local/charts/<profile>/20260507-150912.html
     ```
     Done. Skip every other phase. Don't re-execute SQL. Don't re-render. Don't add 5 follow-ups (this is a UI action, not a fresh answer).
   - **`chart_path` is null** (last query was a 1×1 scalar that didn't render a chart) → surface: "The last answer didn't render a chart (it was a single number). Ask me a new question and I'll generate a fresh report."
   - **`chart_path` set but the file is missing** (user deleted `<artifacts_dir>/local/charts/<profile>/`) → surface: "The chart file is gone — `<path>` no longer exists. Ask me the question again and I'll regenerate it."
   - **Query log empty or missing** → surface: "I don't have any prior queries to reopen. Ask me a question first."

This phase neither logs anything new to `query_log.jsonl` nor sends telemetry — re-opening an existing artifact isn't a query event.

### 2b — assemble the prompt via the examples-first traversal

For a single profile, follow the **examples-first canonical loop** — the subject area is the scoping unit, so you never dump the whole schema. (Cross-profile federation is **2b.federation** below; it's orthogonal to this loop.)

**Step 1 — pick the subject area(s).** `cli areas "$ROOT"` → choose the area(s) whose description matches the question's intent. Most questions touch one area; cross-area ones (a join spanning two areas) select both, and the org's `cross_subject_area_relationships` supply the join.

**Step 2 — examples first (strongest signal).** `cli examples "$ROOT" --area <area> --query "<question>"`. If `high_confidence` is true, mirror the top match's tagged tables / columns / metric / SQL shape and jump to step 5 — skip cold-start resolution.

**Step 3 *(cold start only)* — resolve entities + metrics + opaque literals.** Match the question's terms to the area's entities (and `metrics`). For any opaque literal in the question (an ID-looking token), the CLI entity matching recognizes its type via `value_pattern` (folded into `get_datasource_schema` on the MCP surface, not a separate tool); if it returns `clarify`, ask the user one targeted question rather than guessing.

**Step 4 *(cold start only)* — choose tables + columns** from what resolved (entity `maps_to`, metric `source_tables`).

**Step 5 — compound context fetch.** `cli context "$ROOT" --area <area> --tables … [--columns …]` returns columns (scoped by the area's `expose_column_groups` — wide tables disclose only their exposed groups), `default_filters`, relationships (with cardinality + signers), `caveats`, `value_transforms`, and metrics, in one round-trip.

**Step 6 — assemble the generator prompt** in this order, then produce ONE SQL statement (first statement only if several are emitted):

1. **System** — "Write one valid SQL statement for `<DB_TYPE>` (ANSI_SQL + `<DB_TYPE>` tweaks per dialect-rules.md). Output ONLY SQL. Prefer indexed/`recommended_filters` columns on large tables. Apply each column's `value_transform` when selecting/filtering it. **A `sensitive` column is the model author asking for care, not a locked door.** Prefer using it in `COUNT`/`COUNT(DISTINCT …)`, `GROUP BY`, `WHERE`, and `JOIN` over projecting its raw per-row values. So: (a) 'how many unique customer emails?' → `SELECT COUNT(DISTINCT email)` and report the count. (b) To disambiguate identical display labels (two customers with the same name), put the entity's **non-sensitive key (`id`)** in the output rather than the raw email/phone. (c) When the question genuinely needs the values — 'I need the mailing list' — project them and **say in the answer that you did**; the receipt records it too. Nothing refuses this, so the care is yours to exercise and yours to be transparent about. A column that must never be readable is not in the model at all, and any statement naming it is refused as out of scope. Use a metric's `bindings` SQL VERBATIM when the question names that metric (or a synonym)."
2. **Schema context** — the `get_table_context` output for the chosen tables (columns + types + caveats + value_transforms), the area's relationships (rendered as `from.col → to.col [cardinality]`), and the area's metrics (`<name>: <binding> -- <calculation>` + synonyms). **`default_filters` ARE yours to apply.** Nothing applies them for you — but the receipt DOES report whether you did, per table reference, so an omission is visible to the user in the report beside the answer. If a table declares one and the question does not deliberately ask about the rows it excludes, write it into the `WHERE` clause. Two things to watch, because `get_table_context` has already rewritten them: the `{alias}` placeholder is gone — each filter comes back qualified with the **bare table name**, so re-qualify it to whatever alias you actually used, or the database rejects the statement. And a filter that still carries a `:param` marker (e.g. `orders.tenant_id = :tenant_id`) has no value to bind — leave it OUT and say so in the answer rather than emitting SQL that won't parse. Also DO honor any caveats.

   **Unreviewed metrics are USED, not refused.** When the question names a metric whose `review_state ≠ approved`, still use its binding and answer — do NOT block or refuse on it. The trust layer surfaces it **on the receipt**, not as a hard gate: the metric rides on the `receipt.columns.items` entry for the output column that computes it (`kind: "output"`, `status: "matched"`) carrying its own `review_state`, and the report's approve/change banner is driven off exactly that field (Phase 4e.iii.5). The loader already drops only `rejected` metrics; an `unreviewed`/`proposed` one is yours to use, with the surfaced review state carrying the honesty. (Same for unreviewed joins/entities and `stale` entries — surface, never refuse.)
3. **Datasource context** — `datasource.md` (step 1d.2), heading `## Datasource context`. Binding domain context.
4. **User memory** — `USER_MEMORY.md` (step 1d.1), heading `## User memory (preferences and policies)`.
5. **Few-shot examples** — the ranked matches from step 2.
6. **User question.**

**The fan/chasm safety pass runs as a pre-execution step, on every tier.** Before you execute the generated SQL, pass it through `sm prepare "$ROOT" --area <area> --sql-file <path>` (Phase 3a). It runs the fan-trap / chasm-trap pre-flight and the aggregation-semantics checks, and returns the SQL to actually run, which is always the SQL you gave it: this command never rewrites your statement and never refuses it. What it returns alongside is `findings` — see Phase 3a for how to act on them. Tier-independent — works whether you execute via psql, the Python driver, or DuckDB — so the checks never depend on the execution path. It does **not** apply the area's `default_filters`, and nothing else does either: those are declarative only, so put any you need into the statement yourself at step 6, and `sm prepare` emits no `applied_filters` key of its own (an always-empty list would read as "we checked, none applied"). Which ones your finished statement satisfied is settled later, by `sm receipt`, on `tables.items[].filters` (Phase 4e.iii.5). The findings DO ride on the receipt too, in its `aggregates` section, so the panel draws them beside the answer.

#### 2b.federation — cross-database queries

When the question references datasets from ≥ 2 different profiles (e.g., ITSM in Redshift × finance in MySQL), the skill routes the SQL through DuckDB, which ATTACHes both databases in one session and runs a native federated JOIN.

**Detecting federation.** Extend Pass 1 of the two-pass retrieval to pick `(profile, schema, table)` tuples instead of just `(schema, table)`. If the picked set spans `len({tuple.profile}) > 1`, federation mode is on.

For small databases (under 50 tables), build the union of every profile's index up front and run Pass 1 against that combined index — the picker then chooses across profiles automatically. For larger setups Pass 1 already runs; just include `profile` in each entry.

The Pass 1 prompt loads `<artifacts_dir>/local/cross_profile_relationships.yaml` (if present) so the picker knows about declared cross-profile JOIN paths. If the file is missing, the picker falls back to inferring relationships from column-name/type matching across profile indexes — best-effort, with a warning to the user that confidence is lower.

**`<artifacts_dir>/local/cross_profile_relationships.yaml`** (optional) — declares known JOIN paths across profiles:

```yaml
version: "0.1.1"
relationships:
  - name: itsm_assets_to_finance_cost_centers
    from_profile: itsm
    from_dataset: public.assets
    from_columns: [department_id]
    to_profile: finance
    to_dataset: dbo.cost_centers
    to_columns: [dept_id]
    description: ITSM assets carry the same dept_id as finance cost centers.
```

Loaded at session start the same way per-profile indexes are loaded.

**Building the federated SQL.** When federation mode is active, the schema-context section of the prompt uses **three-part dataset names** matching the DuckDB ATTACH alias: `<profile>.<schema>.<table>` (e.g., `itsm.public.assets`, `finance.dbo.cost_centers`). Cross-profile relationships from `cross_profile_relationships.yaml` are rendered alongside per-profile relationships. The model produces SQL using these three-part names.

**Verifying DuckDB is available.** Look up `tool_paths.duckdb` from `<artifacts_dir>/local/.config`. If missing, surface:

```
Cross-database queries need DuckDB. Install it with `brew install duckdb`
(or apt / download) and re-run.
```

…and stop.

**Verifying credentials are set up for every profile.** For each profile in the picked set, check that the corresponding auth file exists (`<artifacts_dir>/local/.pgpass`, `<artifacts_dir>/local/.mysql.cnf`). Missing → run `python3 "$AGAMI_PLUGIN_ROOT/scripts/setup_pgauth.py" --profile <profile>` for each gap, then re-check.

**Generating the temp init file.**

```bash
init_file=$(python3 "$AGAMI_PLUGIN_ROOT/scripts/build_duckdb_attach.py" \
  --profiles "$P1" "$P2")
# init_file is the path of a chmod-600 file in <artifacts_dir>/local/.duckdb_init_*.sql.
# Credentials are inside that file, NOT on the command line.
```

**Running the SQL.**

```bash
duckdb -init "$init_file" -c "$FEDERATED_SQL" --csv
```

The visible Bash command shows only the path — DuckDB reads the ATTACH credentials silently from the init file.

**Tear-down.** After the query completes (success or failure), delete the init file:

```bash
rm -f "$init_file"
```

The next invocation also self-cleans any `.duckdb_init_*.sql` older than 1 hour in case a prior run crashed:

```bash
find "<artifacts_dir>/local" -maxdepth 1 -name '.duckdb_init_*.sql' -mmin +60 -delete 2>/dev/null
```

**Performance warning.** Federated joins through DuckDB scanners are bounded by network round-trips. If both sides of the join estimate to > 100k rows, surface a one-liner before running:

> This federated query may take 30–120s (network round-trips for `<P1>` × `<P2>`). Want to tighten the filter first?

Options: `Run anyway (Recommended for one-off)` / `Let me add a filter` / `Cancel`.

**Type alignment.** Postgres `numeric(10,2)` joined with MySQL `decimal(10,2)` works. Mismatched types (date vs string, integer vs uuid) need an explicit `CAST` in the generated SQL. The Phase 2b prompt instructs the LLM about this:

> When joining across profiles, prefer explicit `CAST(<col> AS <type>)` for any pair where the types might differ (e.g., timestamps stored as strings on one side, dates on the other).

**No Snowflake federation.** DuckDB's `snowflake_scanner` is experimental and not packaged with the standard binary. If a profile in the picked set has `db_type=snowflake`, `build_duckdb_attach.py` exits with a clear error: surface it to the user and suggest pre-aggregating one side as a CSV.

### 2c — safety checks

Apply [`shared/sql-generation-rules.md`](../../shared/sql-generation-rules.md):

- **No DDL/DML.** Refuse on `DROP`, `DELETE`, `INSERT`, `UPDATE`, `ALTER`, `TRUNCATE`, `CREATE`, `GRANT`, `REVOKE`. Regenerate with explicit "SELECT only" framing.
- **No system tables.** Refuse on `pg_catalog`, `information_schema`, `mysql.*`, `sys.*` unless the user is explicitly asking about schema metadata.
- **NULL-safe division** via `NULLIF(denominator, 0)`.
- **`agami.type` consistency** — if the SQL applies a numeric aggregate (`SUM`, `AVG`) to a field whose `agami.type` is `string` or `boolean`, refuse and regenerate. Type info exists for a reason.

### 2d — risk assessment + time estimate for large tables

For each dataset touched by the SQL, look up its `agami.performance_hints`:

`recommended_filters` is a **list of column names** (introspection seeds it with a large table's date/time columns — the columns worth filtering on to avoid a full scan). Check whether the generated SQL's WHERE filters on **any** of them.

- `estimated_row_count > 1_000_000` AND the WHERE filters on **none** of the table's `recommended_filters`:
  → **HIGH risk**. Surface a banner before executing — **name a suggested column** when one exists: "This query scans `<dataset>` (~<row_count>) with no filter on `<recommended_filters[0]>`. Estimated time: <est>. Narrow it — e.g. a date range on `<recommended_filters[0]>` — or proceed anyway?" AskUserQuestion: `Add a filter` / `Proceed anyway` / `Cancel`. If `recommended_filters` is **empty** (no known good filter for this table), use the generic "…without a filter. Add one, or proceed?" wording.
- `100k–1M` rows with no filter on a `recommended_filters` column → **MEDIUM**. Note in response footer; proceed.
- A query that **does** filter on a `recommended_filters` column → treat as narrowed: drop a risk tier (don't HIGH-warn just because the table is big). Otherwise → **LOW**. Proceed silently.

**Time estimate (announced BEFORE Phase 3 execution).** Long-running queries kill the user's confidence — they don't know if the skill is hung or actually working. Before running any non-LOW query, surface a one-liner with the rough wall-clock estimate so they can wait without anxiety:

```
Running this against ~12M rows in <dataset> — estimated 30–90s. I'll narrate when results land.
```

Estimation table (rough, calibrated to common Postgres / Snowflake shapes — adjust as needed from the latency log over time):

| Largest scanned dataset | With indexed filter (`WHERE` matches `agami.performance_hints.indexes`) | Without indexed filter (full scan) |
|---|---|---|
| < 100k rows | < 1s | < 2s |
| 100k–1M | 1–5s | 5–30s |
| 1M–10M | 5–15s | 30–120s |
| 10M–100M | 15–60s | **2–10 min** — ALWAYS warn even if filter is present |
| > 100M | 30–120s | **> 10 min** — block as HIGH risk; offer to add filter or sample |

Snowflake-specific: add 5–30s on top of any estimate for warehouse spin-up if the warehouse has been idle (the query log can detect "first query in this session" → assume cold). Federation (Phase 2b.federation) doubles or triples estimates due to network round-trips — surface "this federated query may take 30–120s" before running, regardless of estimated_row_count.

If the estimate exceeds 30s, also surface: "Cancel anytime — Ctrl+C in CLI, or just send another message." The user should know they're not trapped waiting.

---

## Phase 3: Execute

### HARD RULE — never paste raw output in chat

The Bash result (CSV stdout, stderr, exit code) is for the skill to parse, not for the user to read. **Never paste the raw CSV / TSV from the Bash result into the assistant's response text.** No "Here's what came back: …", no markdown code-fence dumps of the result. Parse internally, then surface the polished output per Phase 4. The host shows the Bash tool call as a collapsible — that's enough provenance for users who want to dig.

### 3a — safety pass, then run the SQL

**Step 1 — prepare (every tier).** Write the generated SQL to a temp file, then run the tier-independent safety pass:

```bash
bash "$AGAMI_PLUGIN_ROOT/scripts/sm" prepare "$ROOT" --area <area> --sql-file /tmp/agami-q.sql
```

It returns JSON: `{sql, findings, units}`, and it **always exits 0**. The returned `sql` is the statement you handed in, byte for byte — this command never rewrites and never refuses. (`units` is the `{output_column: unit}` map traced through that statement — keep it for the table render in 4d.)

`findings` is a list, usually empty. Each entry is `{risk, reason, triggering_joins}` and each one is a **fact about the statement**, not a verdict on it:

- `fan_trap` / `chasm_trap` — a join multiplies the rows an aggregate is computed from. `triggering_joins` names which join does it.
- `fan_out_invariant` — the same multiplication, on an aggregate it cannot move (`MIN`, `MAX`, a `DISTINCT` one, `BOOL_AND` / `BOOL_OR`). The rows were duplicated; the number is the same either way. `triggering_joins` names the join, exactly as above.
- `bad_aggregation` — a `SUM` of a rate or an identifier, or an `AVG` of one.
- `semi_additive` — a `SUM` of a balance across a time grain, which multiplies a stock.

**A finding is not a refusal, and it is not automatically a bug.** Whether a multiplied total is wrong depends on what was asked: the same statement is wrong for order revenue and right for line-item exposure. The pre-flight does not have the question. You do. So this is the point where the judgement gets made, out loud, by you:

- **The finding is not what the user meant** — restructure and re-prepare. Typically: pre-aggregate each measure in its own CTE and outer-join them (chasm), move the aggregate into a window function to keep raw rows, or drop a join the query does not otherwise use (fan, where the many side is touched nowhere but the `ON` clause). **Say in the answer** that you restructured to avoid a `<risk>`. The restructured statement carries no finding, so the receipt says nothing about it — that sentence is yours to write, and it is what separates this from a guard silently swapping the statement underneath the user.
- **The finding IS what the user meant** — run it and say so. "This counts each order once per line item, which is what 'line-item exposure' asks for." A correct answer to the question asked is not a defect, and refusing it would have been the wrong call.
- **You cannot tell which** — the fan-out join also filters or groups, so the candidate readings return *different numbers* (e.g. "loans with ≥1 payment since January" versus a payment-weighted total). Do **not** pick one. Ask: a short "Did you mean…?" with 2–3 concrete interpretations, one plain-language sentence each, no SQL. Generate for the one they choose.

Either way the finding rides on the answer's receipt, in the `aggregates` section, so a user reading the result can see what was found without taking your word for it.

**An empty `findings` is not by itself a clean bill of health.** Check `unchecked` alongside it: it is `null` when the checks ran, and a sentence when they could not — sqlglot missing, the statement unparseable, or no SELECT in it. All three yield the same empty list a genuinely clean statement does, so a non-null `unchecked` means you have learned nothing about this statement's aggregates and should say so rather than implying it passed.

Even when `unchecked` is null, "found nothing" is bounded. The receipt's `aggregates` marker states what the checks do not reach — an aggregate inside a CTE or a subquery, one in `HAVING` or `ORDER BY` — and you should surface that marker rather than presenting an empty section as clean.

**Step 2 — execute the returned `sql`** via the tier's tool from [`shared/connection-reference.md → CLI Connection Commands`](../../shared/connection-reference.md#cli-connection-commands) — psql / mysql / snowsql / sqlite3 / DuckDB, or the Python driver. (If you use `python -m execute_sql`, do **NOT** pass `--no-safety`. `sm prepare` runs the fan/chasm and aggregation checks, which report; it does **not** run the table-scope, `SELECT *` or column-scope gates, which refuse. Those are the enforcement, they live only in `execute_sql`, and skipping them lets a hallucinated table or column reach the warehouse. The checks that would be doubled are cheap; the gates that would be skipped are not optional.) Wrap in a high-resolution timer; capture stdout (CSV rows), stderr (errors), exit code. Route a non-zero exit through the error classifier (Phase 3b).

### 3b — error handling + auto-retry

Route any non-zero exit through [`shared/db_error_classifier.md`](../../shared/db_error_classifier.md). Behavior per kind:

| `error_kind` | Behavior |
|---|---|
| `auth`, `dsn`, `network` | Stop. Surface the one-line remediation. No retry. |
| `driver_missing` | Fall through to the next available method (native CLI → DuckDB → Python driver). |
| `permission` | Stop. DB user lacks SELECT on the touched dataset. |
| `sign_in_required` | Stop. The person's own sign-in expired or is missing — relay the message (sign in again, then start a new conversation). Nothing is broken; don't retry. |
| `column_not_found`, `table_not_found`, `syntax` | Auto-retry up to **2** times. Pass the error back to the SQL generator: "The previous SQL failed with `<one-line classifier message>`. Regenerate using only table / column names from the schema context above." |
| `other` | Stop. Surface raw error truncated to 200 chars. |

A statement stopped for taking too long no longer arrives here as an error. A per-statement deadline
is a **`resource_limit` refusal**, which carries its own remediation because it is a decision the
server made — handle it as a refusal, not as a classified failure. The `timeout` failure kind now
has one producer, the supervisor stopping an executor that never responded, and there is nothing
query-specific to suggest for it.

After 2 retries with no success, stop. Don't loop.

### 3c — parse rows

Parse the CSV stdout. Header row = column names. Body rows = data.

**Sanitize column headers before display.** SQL aliasing slips happen — bare `n`, `cnt`, single-letter columns, `?column?` (Postgres unaliased), and shouty all-uppercase Snowflake names are common. Apply this header transform **once** before any rendering surface (4d markdown, 4e HTML, chart axis labels):

| Raw header | Display header | Why |
|---|---|---|
| `n`, `N`, `cnt`, `CNT` | `Count` | Bare counts are unreadable; "Count" is universal |
| `?column?` (Postgres unaliased), `_col0`, `_col1` | Drop the column entirely from the visible output (parse warning to log: "unaliased column in result; SQL generator should always alias") |
| Single uppercase word (Snowflake default — `STATUS`, `AMOUNT`) | `Status`, `Amount` (title-case the word, lowercase trailing letters) |
| Snake_case (`order_count`, `customer_name`) | Title-case with spaces (`Order Count`, `Customer Name`) — except when ending in a unit (`_amount` → `_amount` raw is OK if unit shows in the header parens, like `Avg Outstanding (INR)`) |
| Already title-case or sentence-case | Leave as-is |

The header transform is purely cosmetic — the underlying alias stays in the SQL and the `tables_used` log. Don't rename the column in the result data; only its rendered label.

Format every cell per its column's `type` and **`unit`** (both come from `get_table_context` — `unit` is the structured currency/unit set during onboarding's currency ask or by a correction). For a metric result column, use the **metric's** `unit`. The canonical mapping is `semantic_model/units.py` (`format_value`) — the table below mirrors it; when in doubt, match it. The same formatting applies to **every** downstream surface — chat markdown table (4d), HTML `table_rows` (4e.iii), AND chart `labels`. **Also pass the value column's `unit` into each chart section's `"unit"` field** so the chart's y-axis + tooltips format the symbol + grouping deterministically (the chart template applies it; you don't hand-format chart axes — and the `datasets` data stays RAW numbers).

**Numbers are formatted by code, not by you.** The chat markdown table (4d) is rendered by `sm format-table` and the chart by the template — both via `units.py`, in full and exact (a verification surface; never abbreviate or round). Your job in 3c is the **non-numeric** display (header sanitization, dates → `MMM D, YYYY`, booleans → Yes/No, `choice_field` → label) and supplying each column's **`unit`**; the numeric cells are then emitted deterministically downstream. The type/unit table below is the reference `units.py` implements.

| `type` | `unit` | Format |
|---|---|---|
| `decimal` / `integer` | `usd`, `dollars` | `$148.95` (always show `$`, 2 decimals, locale grouping) |
| `decimal` / `integer` | `eur` | `€148.95` |
| `decimal` / `integer` | `gbp` | `£148.95` |
| `decimal` / `integer` | `jpy` | `¥148` (no decimals — JPY has no minor unit) |
| `decimal` / `integer` | `inr` | `₹2,162,087` (Indian Rupee symbol; for amounts > 100k optionally use `2.16 Cr` / `21.6 L` if USER_MEMORY says "use Indian numbering") |
| `decimal` / `integer` | `cad` | `CA$148.95` (or `$148.95` if locale is Canadian) |
| `decimal` / `integer` | `aud` | `A$148.95` |
| `decimal` / `integer` | `chf` | `CHF 148.95` (no symbol convention; show code + space) |
| `decimal` / `integer` | other ISO 4217 code | `<UPPERCASE_CODE> <number>` — show the code as a prefix (`SEK 148.95`, `BRL 148.95`). Never strip the unit silently. |
| `decimal` | `percent` | `12.4%` (1 decimal) |
| `decimal` | (other / none) | `1,234.56` (commas, 2 decimals) |
| `integer` | (no currency unit) | `1,234` (commas, no decimals) |
| `date` | — | **`May 6, 2026`** (`MMM D, YYYY`) — never raw ISO `2026-05-06`, never epoch numbers |
| `timestamp` | — | **`May 6, 2026 3:14 PM`** (`MMM D, YYYY h:mm A`) — never raw ISO with `T` separator, never microsecond precision |
| `integer` / `string` | `epoch_s`/`epoch_ms`/`epoch_us`/`epoch_ns`, `yyyymmdd` | a **date encoding** — `format-table` renders it human-readably (epoch → `YYYY-MM-DD HH:MM:SS UTC`) deterministically; don't hand-format. **In the SQL, convert it** (`to_timestamp(created_ts)` for epoch_s, `to_timestamp(created_ts/1000)` for ms, etc.) so filters/grouping work and the result is a real date. |
| `boolean` | — | `Yes` / `No` (or `true` / `false` if the user prefers — read USER_MEMORY) |
| `string` (with `choice_field`) | — | the choice's display label, not the stored value |
| `string` (other) | — | as-is |

**Timezone:** when a date/timestamp result column has a `timezone` in `get_table_context` (e.g. `UTC` for epoch columns, `offset-aware` for a TZ-aware timestamp), **state it once in the prose insight** ("times shown in UTC") so the reader isn't guessing — and especially flag it when the stored value is a naive/`offset-aware` timestamp whose zone differs from the user's. Epoch columns are UTC by definition; `format-table` already labels each rendered cell `… UTC`.

**Hard rule for currency:** if the field has a currency `agami.unit`, show the symbol or code in **every** cell value — never omit it because the column header already mentions the currency. The user's screenshot showed bare `2,162,087` in an "Avg Outstanding (INR)" column; the cell should read `₹2,162,087`. Headers can drop redundant unit (the column header `Avg Outstanding` is enough when every cell is `₹...`), but cells should always carry the symbol so a screenshot or copy-paste of a single cell stays unambiguous.

**Hard rule for dates: never display ISO timestamps like `2026-05-07T15:14:00.000Z`, epoch seconds, or any other machine-format string.** Cell values OR chart-axis labels — both must be human-readable. If the column shows months only (typical for time-series charts grouped by month), use `MMM YYYY` (e.g. `May 2026`); for daily granularity use `MMM D` if all data is in the current year, else `MMM D, YYYY`. The skill is responsible for inferring the right grain from the data.

If the user has stated a date-format preference in `<artifacts_dir>/USER_MEMORY.md` (e.g. "use ISO dates" or "use DD/MM/YYYY"), respect that. The defaults above apply only when USER_MEMORY is silent.

If row count > 30:
- The chat preview shows the first **30** rows only — markdown tables past ~30 rows scroll forever and bury the insight.
- The HTML report (Phase 4e) contains the full set in a paginated table.
- For row counts > 30, **auto-write the CSV to `<artifacts_dir>/local/exports/<profile>/<ts>.csv`** at the same time as the HTML report, without waiting for the user to ask. Surface both paths in Phase 4d's footer.
- Footer line under the table: `Showing first 30 of <N> rows · full set in CSV: <csv-path> · HTML report: <html-path>`. Use thousands separators on `<N>` (e.g. `4,213`).
- If `<N>` is huge (> 100k), additionally suggest tightening the filter inline: "If you want to slice by region or date, say so and I'll re-run."

CSVs open natively in Excel / Numbers / Google Sheets, so "export to Excel" routes to this same CSV path — no separate `.xlsx` flow in v1.

If row count == 0:
- "No rows matched. The query was: …" (show SQL).
- Suggest a relaxation if applicable.

---

## Phase 4: Present

The chat reply follows this **strict order**, with NO other content interleaved (no SQL, no Bash output, no "let me know if…" filler):

  4a Approach → 4b Fetching → 4c Insight → 4d Table → 4e HTML report path → 4f Numbered follow-ups → (optional 4g CSV path)

Each step is short. The whole reply should fit in a typical chat viewport without scrolling.

### 4a — Approach (one sentence, plain English)

Open with a single sentence describing **how** you'll answer, in plain English. No SQL keywords (`SELECT`, `JOIN`, `GROUP BY`). No table or column names — describe the dimensions in user-language. Examples:

- "I'll group orders by status across the last 30 days and rank by count."
- "I'll pull total spend per customer and sort to find the top 5."
- "I'll compare this month's revenue to last month's, broken down by region."

For multi-section reports (broad questions), the approach sentence describes the narrative across all sections: "I'll cover four angles — revenue trend, top customers, order status, and region split."

### 4b — Fetching (one sentence, counts only)

Right after the approach, one sentence about the data we just pulled, in counts and dimensions. Examples:

- "Pulled 6 rows across 4 statuses."
- "Pulled 247 orders spanning the last 30 days."
- "Pulled 12 monthly buckets with revenue totals."

For multi-section: "Pulled data for 4 sections (12 + 5 + 6 + 3 rows)."

### 4c — Insight first

One sentence stating the answer. Lead with the most surprising or actionable finding. Examples:

- "**Carol Chen** is the top spender at **$148.95** — about 3x the next customer."
- "Revenue is up **12% MoM**; the surge came from EU customers."
- "60% of orders shipped on time this month, up from 48% last month."

For multi-section: a 1–3 sentence executive summary across all sections (the same summary that goes into `REPORT_SUMMARY` for the HTML).

### 4d — Markdown table (single-section reports only)

**Render the table deterministically — do NOT hand-type the number cells.** This is a verification surface; an exact number is mandatory (no rounding, no `1.2L`/`2.16Cr` abbreviation, no dropped decimals). Pass the result CSV (first 30 rows, headers sanitized per 3c) and the **`units` map that `sm prepare` already returned in 3a** through the packaged formatter, then **embed its output verbatim**:
```bash
bash "$AGAMI_PLUGIN_ROOT/scripts/sm" format-table --csv-file /tmp/agami-result.csv --units "$PREPARE_UNITS"
```
`prepare`'s `units` is keyed by output column and **traced through the SQL** — so `SUM(amount) AS total_outstanding` correctly carries amount's currency (a bare name match would miss the summed total). `format-table` then formats every numeric cell in full via `units.py` (symbol + Indian/western grouping, exact decimals) and passes non-numbers through. Same map + formatter the **MCP** uses, so verification numbers are identical regardless of host/LLM. (If a header was prettified in 3c, key the units to the prettified name.) Wide tables (> 8 cols) → vertical layout + "wide table — see HTML for the full grid".

**Cap the chat preview at 30 rows** per Phase 3c — even when the user asked for "all leads with credit rating > 700" and the result is 4,213 rows, the chat shows the first 30 and points them at the CSV + HTML report. The full set lives in the artifacts on disk. The footer line ("Showing first 30 of 4,213 rows · full set in CSV: …") is the contract that tells the user where to find everything.

**Multi-section reports skip the table in chat.** The chat already has the insight; the per-section tables live in the HTML report. Multi-section chat output is: approach + fetching + summary + a short bulleted list of section titles + HTML path + 5 follow-ups. No tables in chat.

### 4e — Build ONE coherent HTML report (one file, N sections)

The output is **one self-contained HTML file** at `<artifacts_dir>/local/charts/<profile>/<ts>.html`, no matter how broad the question is. Broad questions decompose into multiple sub-questions; each sub-question becomes a **section** inside the same file. Each section has its own chart + table + insight + SQL. **Never write multiple HTML files for one user question. Never open multiple browser tabs.**

Skip the report only when the result is a single 1×1 scalar (e.g., `SELECT COUNT(*) FROM orders` returning `42`) — for those, the chat answer is enough.

#### 4e.i — decompose the question into sections

If the user asked something narrow ("top 5 customers by spend"), produce **one** section. Done.

If the user asked something broad ("how is the business doing", "tell me about our customers", "how did we do last quarter"), break it into **2–5 sub-questions** that together tell a narrative. Pick the dimensions that matter for that schema. Examples:

- "How is the business doing?" →
  1. Revenue trend over the last 12 months
  2. Top 5 customers by spend this quarter
  3. Order count by status this quarter
  4. Top 5 products by revenue this quarter

- "Tell me about our customers" →
  1. Customer count by region
  2. Top 10 customers by lifetime spend
  3. New customers per month
  4. Active vs inactive split

Choose sub-questions that:
- Each map to ONE SQL query
- Each return a result shape that produces a useful chart (or a small table when no chart applies)
- Don't repeat the same data sliced differently — pick distinct angles
- Are bounded — never more than 5 sections in v1; if the schema invites more, ship the top 4–5 and add a "What else can I look at?" follow-up

When in doubt about how broad the user wants to go, ask via AskUserQuestion before generating: "I can answer this as a focused query or build a 4-section report. Which?"

#### 4e.ii — pick a chart type per section

For each section's SQL result, read `agami.type` for each result column and pick:

| Result shape | `chart_type` |
|---|---|
| 1 categorical (`string`) + 1 numeric | `bar` (use `pie` / `doughnut` if ≤ 6 categories) |
| 1 time (`timestamp` / `date`) + 1+ numeric | `line` |
| 2 numeric | `scatter` |
| 1 categorical + multiple numeric | grouped `bar` (still `bar`) |
| Categorical-only / single-column / 1×1 scalar | `null` — section still renders without a chart |

If the user override-says `--chart pie|line|...` for the **whole** report, apply it to every section that supports a chart.

#### 4e.iii — build the SECTIONS file via `csv_to_sections.py` (do NOT hand-write the numbers)

**You do NOT transcribe result numbers into JSON.** `csv_to_sections.py` reads each section's result CSV + the `units` map (from `sm prepare`) and builds the section's `labels`, `table_rows` (formatted exactly via `units.py`), and `datasets[].data` (raw numbers) — so the chart can never disagree with the table, and a miscopy is impossible. You supply only the **presentation spec**.

Write a spec file `/tmp/agami-spec-<ts>.json` — a JSON array, one object per section:

```json
[{
  "title":       "<sub-question / heading>",
  "insights":    "<1-3 sentence plain-English insight>",
  "chart_type":  "bar|line|pie|doughnut|scatter|null",
  "csv_file":    "/tmp/agami-result-<n>.csv",
  "units":       { },
  "sql_file":    "/tmp/agami-q-<n>.sql",
  "label_col":   0,
  "value_cols":  [1],
  "header_relabels": {"total_amount": "Total Spend"}
}]
```

Your fields: `title` + `insights` (prose), `chart_type`, `label_col` (which column is the x-axis/label, default 0), `value_cols` (chart value columns, default: all non-label), and optional cosmetic `header_relabels`. Everything numeric is the script's job — keep each section's result CSV (Phase 3) and its `sql_file` so the spec can point at them. `units` is the object `sm prepare` returned. When `chart_type` is `null` the chart card is skipped; `table_rows` still render.

```bash
python3 "$AGAMI_PLUGIN_ROOT/scripts/csv_to_sections.py" \
  --spec /tmp/agami-spec-<ts>.json --out /tmp/agami-sections-<ts>.json
```

It prints `{ok, data:{section_count}, anomalies}`. **Read `anomalies`** and mention/act on them if they matter (e.g. a value column that isn't numeric → left out of the chart; a query that returned 0 rows) — they're not fatal. The `--out` sections file is what `render_chart.py` reads (4e.iv).

Notes the script handles for you: a time-bucket `label_col` is formatted from its unit/date encoding; a categorical label uses the value as-is; the section's chart-axis `unit` is set automatically only when **all** value columns share one unit (so a currency col plotted next to a count never mis-formats); the `sql` is read verbatim from `sql_file` (no paraphrase, no `...`).

#### 4e.iii.5 — build the trust receipt via `sm receipt` (do NOT hand-build it)

The receipt documents provenance: the tables the statement touched, the joins and metrics it used, the columns it referenced, the assumptions it leaned on, and the model-version pin. **It is assembled deterministically from the SQL + the model** by `runtime.assemble_receipt` — the SAME builder the MCP server uses — so it never depends on you extracting fields by hand:

```bash
bash "$AGAMI_PLUGIN_ROOT/scripts/sm" receipt "$ROOT" \
  --sql-file /tmp/agami-q-<n>.sql \
  > /tmp/agami-receipt-<ts>.json
```

It emits `model_version` plus **five sections** — `columns`, `tables`, `joins`, `aggregates`, `assumptions` — each an object of the form `{"items": [...], "undetermined": "<sentence>" | null}`. It parses the FROM/JOIN scope and matches model entries itself:

| section | what `items[]` holds |
| --- | --- |
| `columns` | every column the statement referenced, as `{"column": "<schema>.<table>.<col>", "metric": null}` — **plus** each model metric whose binding SQL appears in the statement, as `{"column": null, "metric": {...}}`. A metric match is a statement-level fact today, so it has no owning column. The metric object carries `name`, `area`, `definition_prose`, `expression`, `confidence`, `review_state`, `origin`, `signed_off_by`/`_role`/`_at` |
| `tables` | **one entry per REFERENCE, not per table** — a table read twice appears twice. `ref` (the name as the statement wrote it), `alias`, `qname` (what the model resolved it to), `declared`, `rows`, `rows_as_of`, `freshness`, `scope`, `filters`. An **undeclared** reference (a CTE name, say) has `declared: false` and null `qname`/`rows`/`rows_as_of`/`freshness`, because there is no model row for those to be about. `scope` is `main` or `cte:<name>`, each carrying a trailing 1-based `#<n>` when that scope is one of **two or more** arms of a set operation (`UNION` / `INTERSECT` / `EXCEPT`) — `main#1`, `main#2`, `cte:recent#2` — or `subquery`, which **never** takes an ordinal because it names no scope to have arms. A plain `SELECT` and a single-arm CTE body take no suffix either. It is the query scope the reference was written in, and the only thing that tells two entries for the same `ref` apart: a table read in two arms is otherwise two identica