---
plan: true
title: ""
status: iterating
iteration: 2
target_iterations: 8
beads_revision: 0
related_plans: []
created: 2026-02-17
updated: 2026-02-17
---

# Plan: Expose Discussion IDs Across the Read Surface

**Problem**: Agents can't bridge from lore's read output to glab's write API because
`gitlab_discussion_id` — stored in the DB — is never exposed in any output. The read/write
split contract requires lore to emit every identifier an agent needs to construct a glab write
command.

**Scope**: Four workstreams, delivered in order:
1. Add `gitlab_discussion_id` to notes output
2. Add `gitlab_discussion_id` to show command discussion groups
3. Add a standalone `discussions` list command
4. Fix robot-docs to list actual field names instead of opaque type references

---

## Bridge Contract (Cross-Cutting)

Every read payload that surfaces notes or discussions **MUST** include:
- `project_path`
- `noteable_type`
- `parent_iid`
- `gitlab_discussion_id`
- `gitlab_note_id` (when note-level data is returned — i.e., in notes list and show detail)

This contract exists so agents can deterministically construct `glab api` write calls without
cross-referencing multiple commands. Each workstream below must satisfy these fields in its
output.

### Field Filtering Guardrail

In robot mode, `filter_fields` **MUST** force-include Bridge Contract fields even when the
caller passes a narrower `--fields` list. This prevents agents from accidentally stripping
the identifiers they need for write operations.

**Implementation**: Add a `BRIDGE_FIELDS` constant map per entity type. In `filter_fields()`,
when operating in robot mode, union the caller's requested fields with the bridge set before
filtering. Human/table mode keeps existing behavior (no forced fields).

```rust
// src/cli/robot.rs
const BRIDGE_FIELDS_NOTES: &[&str] = &[
    "project_path", "noteable_type", "parent_iid",
    "gitlab_discussion_id", "gitlab_note_id",
];
const BRIDGE_FIELDS_DISCUSSIONS: &[&str] = &[
    "project_path", "noteable_type", "parent_iid",
    "gitlab_discussion_id",
];
```

In `filter_fields`, when entity is `"notes"` or `"discussions"`, merge the bridge set into the
requested fields before filtering the JSON value. This is a ~5-line change to the existing
function.

---

## 1. Add `gitlab_discussion_id` to Notes Output

### Why

The `notes` command joins `discussions` but never SELECTs `d.gitlab_discussion_id`. An agent
that finds a note via `--contains` or `--for-mr` has no way to get the discussion thread ID
needed to reply via `glab api`.

### Current Code

**SQL query** (`src/cli/commands/list.rs:1366-1395`):
```sql
SELECT
    n.id, n.gitlab_id, n.author_username, n.body, n.note_type,
    n.is_system, n.created_at, n.updated_at,
    n.position_new_path, n.position_new_line,
    n.position_old_path, n.position_old_line,
    n.resolvable, n.resolved, n.resolved_by,
    d.noteable_type,
    COALESCE(i.iid, m.iid) AS parent_iid,
    COALESCE(i.title, m.title) AS parent_title,
    p.path_with_namespace AS project_path
FROM notes n
JOIN discussions d ON n.discussion_id = d.id
JOIN projects p ON n.project_id = p.id
LEFT JOIN issues i ON d.issue_id = i.id
LEFT JOIN merge_requests m ON d.merge_request_id = m.id
```

**Row struct** (`src/cli/commands/list.rs:1041-1061`):
```rust
pub struct NoteListRow {
    pub id: i64,
    pub gitlab_id: i64,
    pub author_username: String,
    pub body: Option<String>,
    pub note_type: Option<String>,
    pub is_system: bool,
    pub created_at: i64,
    pub updated_at: i64,
    pub position_new_path: Option<String>,
    pub position_new_line: Option<i64>,
    pub position_old_path: Option<String>,
    pub position_old_line: Option<i64>,
    pub resolvable: bool,
    pub resolved: bool,
    pub resolved_by: Option<String>,
    pub noteable_type: Option<String>,
    pub parent_iid: Option<i64>,
    pub parent_title: Option<String>,
    pub project_path: String,
}
```

**JSON struct** (`src/cli/commands/list.rs:1063-1094`):
```rust
pub struct NoteListRowJson {
    pub id: i64,
    pub gitlab_id: i64,
    pub author_username: String,
    // ... (same fields, no gitlab_discussion_id)
    pub project_path: String,
}
```

### Changes Required

#### 1a. Add field to SQL SELECT

**File**: `src/cli/commands/list.rs` line ~1386

Add `d.gitlab_discussion_id` to the SELECT list. Insert it after
`p.path_with_namespace AS project_path` since it's from the already-joined discussions table.

```sql
-- ADD this line after project_path:
d.gitlab_discussion_id
```

#### 1b. Add field to `NoteListRow` and switch to named column lookup

**File**: `src/cli/commands/list.rs` line ~1060

```rust
pub struct NoteListRow {
    // ... existing fields ...
    pub project_path: String,
    pub gitlab_discussion_id: String,  // ADD
}
```

And in the `query_map` closure (line ~1407), switch from positional indexing to named column
lookup for the new field and ideally all fields. At minimum, the new field uses named lookup
to avoid fragile positional shifts:

```rust
Ok(NoteListRow {
    // ... existing fields using positional gets ...
    project_path: row.get("project_path")?,
    gitlab_discussion_id: row.get("gitlab_discussion_id")?,  // ADD — named lookup
})
```

**Implementation note**: If converting all existing fields from positional to named lookup is
low-risk, do it in this change. The SQL already uses aliases (`AS project_path`, `AS parent_iid`,
etc.) which rusqlite's `row.get("name")` can resolve. This eliminates the fragility of
column-index counting that has caused bugs in the past. If the conversion touches too many
lines, limit named lookup to just the new field and add a follow-up task.

#### 1c. Add field to `NoteListRowJson`

**File**: `src/cli/commands/list.rs` line ~1093

```rust
pub struct NoteListRowJson {
    // ... existing fields ...
    pub project_path: String,
    pub gitlab_discussion_id: String,  // ADD
}
```

And in the `From<&NoteListRow>` impl (line ~1096):

```rust
impl From<&NoteListRow> for NoteListRowJson {
    fn from(row: &NoteListRow) -> Self {
        Self {
            // ... existing fields ...
            project_path: row.project_path.clone(),
            gitlab_discussion_id: row.gitlab_discussion_id.clone(),  // ADD
        }
    }
}
```

#### 1d. Add to CSV header

**File**: `src/cli/commands/list.rs` line ~1004

Add `gitlab_discussion_id` to the CSV header and row output.

#### 1e. Add to table display

**File**: `src/cli/commands/list.rs`, `print_list_notes()` function

Add a column showing a truncated discussion ID (first 8 chars) in the table view.

#### 1f. Update `--fields minimal` preset

**File**: `src/cli/robot.rs` line ~67

```rust
"notes" => ["id", "author_username", "body", "created_at_iso", "gitlab_discussion_id"]
    .iter()
    .map(|s| (*s).to_string())
    .collect(),
```

The discussion ID is critical enough for agent workflows that it belongs in `minimal`.

#### 1g. Add `--gitlab-discussion-id` filter to notes

Allow filtering notes directly by GitLab discussion thread ID (the external string ID, not
the internal integer). This enables one-hop note retrieval from external references — an agent
that received a `gitlab_discussion_id` from another command or webhook can jump straight to
the relevant notes without knowing the internal discussion ID.

**File**: `src/cli/mod.rs` (NotesArgs)

```rust
/// Filter by GitLab discussion ID
#[arg(long, help_heading = "Filters")]
pub gitlab_discussion_id: Option<String>,
```

**File**: `src/cli/commands/list.rs` (NoteListFilters + where clause)

Add `gitlab_discussion_id: Option<String>` to `NoteListFilters`. In the WHERE construction:

```sql
-- When gitlab_discussion_id is provided:
AND d.gitlab_discussion_id = ?
```

This is a single WHERE clause addition — minimal complexity, high value for bridge workflows.

### Tests

**File**: `src/cli/commands/list_tests.rs`

#### Test 1: `gitlab_discussion_id` present in NoteListRowJson

```rust
#[test]
fn note_list_row_json_includes_gitlab_discussion_id() {
    let row = NoteListRow {
        id: 1,
        gitlab_id: 100,
        author_username: "alice".to_string(),
        body: Some("test".to_string()),
        note_type: Some("DiscussionNote".to_string()),
        is_system: false,
        created_at: 1_700_000_000_000,
        updated_at: 1_700_000_000_000,
        position_new_path: None,
        position_new_line: None,
        position_old_path: None,
        position_old_line: None,
        resolvable: false,
        resolved: false,
        resolved_by: None,
        noteable_type: Some("MergeRequest".to_string()),
        parent_iid: Some(42),
        parent_title: Some("Fix bug".to_string()),
        project_path: "group/project".to_string(),
        gitlab_discussion_id: "6a9c1750b37d".to_string(),
    };

    let json_row = NoteListRowJson::from(&row);
    assert_eq!(json_row.gitlab_discussion_id, "6a9c1750b37d");

    let serialized = serde_json::to_value(&json_row).unwrap();
    assert!(serialized.get("gitlab_discussion_id").is_some());
    assert_eq!(
        serialized["gitlab_discussion_id"].as_str().unwrap(),
        "6a9c1750b37d"
    );
}
```

#### Test 2: query_notes returns gitlab_discussion_id from DB

```rust
#[test]
fn query_notes_returns_gitlab_discussion_id() {
    let conn = create_connection(Path::new(":memory:")).unwrap();
    run_migrations(&conn).unwrap();

    // Insert project, discussion, note
    conn.execute(
        "INSERT INTO projects (id, gitlab_project_id, path_with_namespace, web_url)
         VALUES (1, 1, 'group/project', 'https://gitlab.com/group/project')",
        [],
    ).unwrap();
    conn.execute(
        "INSERT INTO discussions (id, gitlab_discussion_id, project_id, issue_id, noteable_type, last_seen_at)
         VALUES (1, 'abc123def456', 1, NULL, 'MergeRequest', 1000)",
        [],
    ).unwrap();
    conn.execute(
        "INSERT INTO merge_requests (id, gitlab_id, project_id, iid, title, state, author_username, source_branch, target_branch, created_at, updated_at, last_seen_at)
         VALUES (1, 1, 1, 99, 'Test MR', 'opened', 'alice', 'feat', 'main', 1000, 1000, 1000)",
        [],
    ).unwrap();
    // Update discussion to reference MR
    conn.execute(
        "UPDATE discussions SET merge_request_id = 1 WHERE id = 1",
        [],
    ).unwrap();
    conn.execute(
        "INSERT INTO notes (id, gitlab_id, discussion_id, project_id, author_username, body, created_at, updated_at, last_seen_at, is_system, position)
         VALUES (1, 500, 1, 1, 'bob', 'test note', 1000, 1000, 1000, 0, 0)",
        [],
    ).unwrap();

    let config = Config::default();
    let filters = NoteListFilters {
        limit: 10,
        project: None,
        author: None,
        note_type: None,
        include_system: false,
        for_issue_iid: None,
        for_mr_iid: None,
        note_id: None,
        gitlab_note_id: None,
        discussion_id: None,
        since: None,
        until: None,
        path: None,
        contains: None,
        resolution: None,
        sort: "created".to_string(),
        order: "desc".to_string(),
    };

    let result = query_notes(&conn, &filters, &config).unwrap();
    assert_eq!(result.notes.len(), 1);
    assert_eq!(result.notes[0].gitlab_discussion_id, "abc123def456");
}
```

#### Test 3: --fields filtering includes gitlab_discussion_id

```rust
#[test]
fn fields_filter_retains_gitlab_discussion_id() {
    let mut value = serde_json::json!({
        "data": {
            "notes": [{
                "id": 1,
                "gitlab_id": 100,
                "author_username": "alice",
                "body": "test",
                "gitlab_discussion_id": "abc123"
            }]
        }
    });

    filter_fields(
        &mut value,
        "notes",
        &["id".to_string(), "gitlab_discussion_id".to_string()],
    );

    let note = &value["data"]["notes"][0];
    assert_eq!(note["id"], 1);
    assert_eq!(note["gitlab_discussion_id"], "abc123");
    assert!(note.get("body").is_none());
}
```

#### Test 4: Bridge fields survive aggressive --fields filtering in robot mode

```rust
#[test]
fn bridge_fields_forced_in_robot_mode() {
    // Agent requests only "body" — bridge fields must still appear
    let mut value = serde_json::json!({
        "data": {
            "notes": [{
                "id": 1,
                "body": "test",
                "project_path": "group/repo",
                "noteable_type": "MergeRequest",
                "parent_iid": 42,
                "gitlab_discussion_id": "abc123",
                "gitlab_note_id": 500
            }]
        }
    });

    // In robot mode, filter_fields merges bridge set
    filter_fields_robot(
        &mut value,
        "notes",
        &["body".to_string()],
    );

    let note = &value["data"]["notes"][0];
    assert_eq!(note["body"], "test");
    // Bridge fields survive despite not being requested:
    assert!(note.get("project_path").is_some());
    assert!(note.get("gitlab_discussion_id").is_some());
    assert!(note.get("parent_iid").is_some());
}
```

#### Test 5: --gitlab-discussion-id filter returns matching notes

```rust
#[test]
fn notes_filter_by_gitlab_discussion_id() {
    let conn = create_test_db();
    // Insert 2 discussions with different gitlab_discussion_ids, each with notes
    // Filter by one gitlab_discussion_id
    // Assert only notes from matching discussion are returned
}
```

---

## 2. Add `gitlab_discussion_id` to Show Command Discussion Groups

### Why

`lore -J issues 42` and `lore -J mrs 99` return discussions grouped by thread, but neither
`DiscussionDetailJson` nor `MrDiscussionDetailJson` includes the `gitlab_discussion_id`. An
agent viewing MR details can see all discussion threads but can't identify which one to reply to.

### Current Code

**Issue discussions** (`src/cli/commands/show.rs:99-102`):
```rust
pub struct DiscussionDetail {
    pub notes: Vec<NoteDetail>,
    pub individual_note: bool,
}
```

**MR discussions** (`src/cli/commands/show.rs:37-40`):
```rust
pub struct MrDiscussionDetail {
    pub notes: Vec<MrNoteDetail>,
    pub individual_note: bool,
}
```

**JSON equivalents** (`show.rs:1001-1003` and `show.rs:1100-1103`):
```rust
pub struct DiscussionDetailJson {
    pub notes: Vec<NoteDetailJson>,
    pub individual_note: bool,
}
pub struct MrDiscussionDetailJson {
    pub notes: Vec<MrNoteDetailJson>,
    pub individual_note: bool,
}
```

**Queries** (`show.rs:325-328` and `show.rs:537-540`):
```sql
SELECT id, individual_note FROM discussions WHERE issue_id = ? ORDER BY first_note_at
SELECT id, individual_note FROM discussions WHERE merge_request_id = ? ORDER BY first_note_at
```

### Changes Required

#### 2a. Add fields to domain structs

**File**: `src/cli/commands/show.rs`

```rust
pub struct DiscussionDetail {
    pub gitlab_discussion_id: String,  // ADD
    pub resolvable: bool,              // ADD — agents need thread state
    pub resolved: bool,                // ADD — agents need thread state
    pub last_note_at: i64,             // ADD — for recency sorting
    pub notes: Vec<NoteDetail>,
    pub individual_note: bool,
}

pub struct MrDiscussionDetail {
    pub gitlab_discussion_id: String,  // ADD
    pub resolvable: bool,              // ADD
    pub resolved: bool,                // ADD
    pub last_note_at: i64,             // ADD
    pub notes: Vec<MrNoteDetail>,
    pub individual_note: bool,
}
```

#### 2b. Add fields to JSON structs

```rust
pub struct DiscussionDetailJson {
    pub gitlab_discussion_id: String,  // ADD
    pub resolvable: bool,              // ADD
    pub resolved: bool,                // ADD
    pub last_note_at_iso: String,      // ADD — ISO formatted
    pub notes: Vec<NoteDetailJson>,
    pub individual_note: bool,
}

pub struct MrDiscussionDetailJson {
    pub gitlab_discussion_id: String,  // ADD
    pub resolvable: bool,              // ADD
    pub resolved: bool,                // ADD
    pub last_note_at_iso: String,      // ADD — ISO formatted
    pub notes: Vec<MrNoteDetailJson>,
    pub individual_note: bool,
}
```

#### 2c. Update queries to SELECT new fields

**Issue discussions** (`show.rs:325`):
```sql
SELECT id, gitlab_discussion_id, individual_note, resolvable, resolved, last_note_at
FROM discussions
WHERE issue_id = ? ORDER BY first_note_at
```

**MR discussions** (`show.rs:537`):
```sql
SELECT id, gitlab_discussion_id, individual_note, resolvable, resolved, last_note_at
FROM discussions
WHERE merge_request_id = ? ORDER BY first_note_at
```

#### 2d. Update query_map closures

The `disc_rows` tuple changes from `(i64, bool)` to a richer shape. Use named columns here
too for clarity:

Issue path (`show.rs:331-335`):
```rust
let disc_rows: Vec<(i64, String, bool, bool, bool, i64)> = disc_stmt
    .query_map([issue_id], |row| {
        Ok((
            row.get("id")?,
            row.get("gitlab_discussion_id")?,
            row.get::<_, i64>("individual_note").map(|v| v == 1)?,
            row.get::<_, i64>("resolvable").map(|v| v == 1)?,
            row.get::<_, i64>("resolved").map(|v| v == 1)?,
            row.get("last_note_at")?,
        ))
    })?
    .collect::<std::result::Result<Vec<_>, _>>()?;
```

And where discussions are constructed (`show.rs:361`):
```rust
for (disc_id, gitlab_disc_id, individual_note, resolvable, resolved, last_note_at) in disc_rows {
    // ... existing note query ...
    discussions.push(DiscussionDetail {
        gitlab_discussion_id: gitlab_disc_id,
        resolvable,
        resolved,
        last_note_at,
        notes,
        individual_note,
    });
}
```

Same pattern for MR discussions (`show.rs:543-560`, `show.rs:598`).

#### 2e. Update From impls

```rust
impl From<&DiscussionDetail> for DiscussionDetailJson {
    fn from(disc: &DiscussionDetail) -> Self {
        Self {
            gitlab_discussion_id: disc.gitlab_discussion_id.clone(),
            resolvable: disc.resolvable,
            resolved: disc.resolved,
            last_note_at_iso: format_iso_timestamp(disc.last_note_at),
            notes: disc.notes.iter().map(|n| n.into()).collect(),
            individual_note: disc.individual_note,
        }
    }
}

impl From<&MrDiscussionDetail> for MrDiscussionDetailJson {
    fn from(disc: &MrDiscussionDetail) -> Self {
        Self {
            gitlab_discussion_id: disc.gitlab_discussion_id.clone(),
            resolvable: disc.resolvable,
            resolved: disc.resolved,
            last_note_at_iso: format_iso_timestamp(disc.last_note_at),
            notes: disc.notes.iter().map(|n| n.into()).collect(),
            individual_note: disc.individual_note,
        }
    }
}
```

#### 2f. Add `gitlab_note_id` to note detail structs in show

While we're here, add `gitlab_id` (as `gitlab_note_id` in JSON) to `NoteDetail`,
`MrNoteDetail`, and their JSON counterparts. Currently show-command notes only have
`author_username`, `body`, `created_at`, `is_system` — no note ID at all, making it impossible
to reference a specific note. This satisfies the Bridge Contract requirement for `gitlab_note_id`
on note-level data.

**Domain structs** — add `gitlab_id: i64` field.
**JSON structs** — add `gitlab_note_id: i64` field.
**Queries** — add `n.gitlab_id` to the note SELECT within show.
**From impls** — map `gitlab_id` → `gitlab_note_id`.

### Tests

**File**: `src/cli/commands/show_tests.rs` (or within show.rs `#[cfg(test)]`)

#### Test 1: Issue show includes gitlab_discussion_id

```rust
#[test]
fn show_issue_includes_gitlab_discussion_id() {
    // Setup: project, issue, discussion with known gitlab_discussion_id, note
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_issue(&conn, 1, 1, 42);
    conn.execute(
        "INSERT INTO discussions (id, gitlab_discussion_id, project_id, issue_id, noteable_type, last_seen_at)
         VALUES (1, 'abc123hex', 1, 1, 'Issue', 1000)",
        [],
    ).unwrap();
    insert_note(&conn, 1, 500, 1, 1, "alice", "hello", false);

    let detail = run_show_issue_with_conn(&conn, 42, None).unwrap();
    assert_eq!(detail.discussions.len(), 1);
    assert_eq!(detail.discussions[0].gitlab_discussion_id, "abc123hex");
}
```

#### Test 2: MR show includes gitlab_discussion_id

Same pattern for MR path.

#### Test 3: JSON serialization includes the field

```rust
#[test]
fn discussion_detail_json_has_gitlab_discussion_id() {
    let detail = DiscussionDetail {
        gitlab_discussion_id: "deadbeef".to_string(),
        resolvable: true,
        resolved: false,
        last_note_at: 1_700_000_000_000,
        notes: vec![],
        individual_note: false,
    };
    let json = DiscussionDetailJson::from(&detail);
    let value = serde_json::to_value(&json).unwrap();
    assert_eq!(value["gitlab_discussion_id"], "deadbeef");
    assert_eq!(value["resolvable"], true);
    assert_eq!(value["resolved"], false);
    assert!(value.get("last_note_at_iso").is_some());
}
```

#### Test 4: Show note includes gitlab_note_id

```rust
#[test]
fn show_note_detail_json_has_gitlab_note_id() {
    // Verify NoteDetailJson serialization includes gitlab_note_id
}
```

---

## 3. Add Standalone `discussions` List Command

### Why

Discussions are a first-class entity in the DB (211K rows) but invisible to agents as a
collection. An agent working on an MR needs to see all threads at a glance — which are
unresolved, who started them, what file they're on — with the `gitlab_discussion_id` needed
to reply. Currently the only way is `lore notes --for-mr 99` which returns flat notes without
discussion grouping.

### Design

```
lore discussions [OPTIONS]

# List all discussions on MR 99 (most common agent use case)
lore -J discussions --for-mr 99

# List unresolved discussions on MR 99
lore -J discussions --for-mr 99 --resolution unresolved

# List discussions on issue 42
lore -J discussions --for-issue 42

# List discussions across a project
lore -J discussions -p group/repo --since 7d

# Look up a specific discussion by GitLab ID
lore -J discussions --gitlab-discussion-id 6a9c1750b37d

# List unresolved threads with latest 2 notes inline (fewer round-trips)
lore -J discussions --for-mr 99 --resolution unresolved --include-notes 2
```

### Response Schema

```json
{
  "ok": true,
  "data": {
    "discussions": [
      {
        "gitlab_discussion_id": "6a9c1750b37d513a...",
        "noteable_type": "MergeRequest",
        "parent_iid": 3929,
        "parent_title": "Resolve \"Switch Health Card\"",
        "project_path": "vs/typescript-code",
        "individual_note": false,
        "note_count": 3,
        "first_author": "elovegrove",
        "first_note_body_snippet": "Ok @teernisse well I really do prefer...",
        "first_note_at_iso": "2026-02-16T14:31:34Z",
        "last_note_at_iso": "2026-02-16T15:02:11Z",
        "resolvable": true,
        "resolved": false,
        "position_new_path": "src/components/SwitchHealthCard.vue",
        "position_new_line": 42,
        "notes": []
      }
    ],
    "total_count": 15,
    "showing": 15
  },
  "meta": { "elapsed_ms": 12 }
}
```

The `notes` array is empty by default (zero overhead). When `--include-notes N` is provided,
each discussion includes up to N of its most recent notes inline. This covers the common
agent pattern of "show me unresolved threads with context" in a single round-trip.

### File Architecture

**No new files.** Follow the existing pattern:
- Args struct: `src/cli/mod.rs` (alongside `NotesArgs`, `IssuesArgs`)
- Query + print functions: `src/cli/commands/list.rs` (alongside `query_notes`, `print_list_notes_json`)
- Handler: `src/main.rs` (alongside `handle_notes`)
- Tests: `src/cli/commands/list_tests.rs`
- Robot-docs: `src/main.rs` robot-docs JSON block

### Implementation Details

#### 3a. CLI Args

**File**: `src/cli/mod.rs`

Add variant to `Commands` enum (after `Notes`):

```rust
/// List discussions
#[command(visible_alias = "discussion")]
Discussions(DiscussionsArgs),
```

Args struct:

```rust
#[derive(Parser)]
pub struct DiscussionsArgs {
    /// Maximum results
    #[arg(short = 'n', long = "limit", default_value = "50", help_heading = "Output")]
    pub limit: usize,

    /// Select output fields (comma-separated, or 'minimal' preset)
    #[arg(long, help_heading = "Output", value_delimiter = ',')]
    pub fields: Option<Vec<String>>,

    /// Output format (table, json, jsonl, csv)
    #[arg(long, default_value = "table", value_parser = ["table", "json", "jsonl", "csv"], help_heading = "Output")]
    pub format: String,

    /// Filter to discussions on a specific issue IID
    #[arg(long, conflicts_with = "for_mr", help_heading = "Filters")]
    pub for_issue: Option<i64>,

    /// Filter to discussions on a specific MR IID
    #[arg(long, conflicts_with = "for_issue", help_heading = "Filters")]
    pub for_mr: Option<i64>,

    /// Filter by project path
    #[arg(short = 'p', long, help_heading = "Filters")]
    pub project: Option<String>,

    /// Filter by GitLab discussion ID
    #[arg(long, help_heading = "Filters")]
    pub gitlab_discussion_id: Option<String>,

    /// Filter by resolution status (unresolved, resolved)
    #[arg(long, value_parser = ["unresolved", "resolved"], help_heading = "Filters")]
    pub resolution: Option<String>,

    /// Filter by time (7d, 2w, 1m, or YYYY-MM-DD)
    #[arg(long, help_heading = "Filters")]
    pub since: Option<String>,

    /// Filter by file path (exact match or prefix with trailing /)
    #[arg(long, help_heading = "Filters")]
    pub path: Option<String>,

    /// Filter by noteable type (Issue, MergeRequest)
    #[arg(long, value_parser = ["Issue", "MergeRequest"], help_heading = "Filters")]
    pub noteable_type: Option<String>,

    /// Include up to N latest notes per discussion (0 = none, default)
    #[arg(long, default_value = "0", help_heading = "Output")]
    pub include_notes: usize,

    /// Sort field (first_note, last_note)
    #[arg(long, value_parser = ["first_note", "last_note"], default_value = "last_note", help_heading = "Sorting")]
    pub sort: String,

    /// Sort ascending (default: descending)
    #[arg(long, help_heading = "Sorting")]
    pub asc: bool,
}
```

#### 3b. Domain Structs

**File**: `src/cli/commands/list.rs`

```rust
#[derive(Debug)]
pub struct DiscussionListRow {
    pub id: i64,
    pub gitlab_discussion_id: String,
    pub noteable_type: String,
    pub parent_iid: Option<i64>,
    pub parent_title: Option<String>,
    pub project_path: String,
    pub individual_note: bool,
    pub note_count: i64,
    pub first_author: Option<String>,
    pub first_note_body: Option<String>,
    pub first_note_at: i64,
    pub last_note_at: i64,
    pub resolvable: bool,
    pub resolved: bool,
    pub position_new_path: Option<String>,
    pub position_new_line: Option<i64>,
}

#[derive(Serialize)]
pub struct DiscussionListRowJson {
    pub gitlab_discussion_id: String,
    pub noteable_type: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub parent_iid: Option<i64>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub parent_title: Option<String>,
    pub project_path: String,
    pub individual_note: bool,
    pub note_count: i64,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub first_author: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub first_note_body_snippet: Option<String>,
    pub first_note_at_iso: String,
    pub last_note_at_iso: String,
    pub resolvable: bool,
    pub resolved: bool,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub position_new_path: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub position_new_line: Option<i64>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub notes: Vec<NoteListRowJson>,
}

pub struct DiscussionListResult {
    pub discussions: Vec<DiscussionListRow>,
    pub total_count: i64,
}

#[derive(Serialize)]
pub struct DiscussionListResultJson {
    pub discussions: Vec<DiscussionListRowJson>,
    pub total_count: i64,
    pub showing: usize,
}
```

The `From` impl truncates `first_note_body` to ~120 chars for the snippet.

The `notes` field on `DiscussionListRowJson` is populated only when `--include-notes N > 0`.
It reuses the existing `NoteListRowJson` struct for consistency — agents get the same note
shape whether they come from `notes`, `show`, or `discussions --include-notes`.

#### 3c. SQL Query

**File**: `src/cli/commands/list.rs`

```rust
pub fn query_discussions(
    conn: &Connection,
    filters: &DiscussionListFilters,
    config: &Config,
) -> Result<DiscussionListResult> {
```

Core query uses a CTE + ranked-notes rollup (window function) to avoid per-row correlated
subqueries. The `ROW_NUMBER()` approach produces a single scan over the notes table, which
is more predictable than repeated LIMIT 1 sub-selects at scale (200K+ discussions):

```sql
WITH filtered_discussions AS (
    SELECT
        d.id, d.gitlab_discussion_id, d.noteable_type, d.project_id,
        d.issue_id, d.merge_request_id, d.individual_note,
        d.first_note_at, d.last_note_at, d.resolvable, d.resolved
    FROM discussions d
    JOIN projects p ON d.project_id = p.id
    {where_sql}
),
ranked_notes AS (
    SELECT
        n.discussion_id,
        n.author_username,
        n.body,
        n.is_system,
        n.position_new_path,
        n.position_new_line,
        ROW_NUMBER() OVER (
            PARTITION BY n.discussion_id
            ORDER BY n.position, n.id
        ) AS rn
    FROM notes n
    WHERE n.discussion_id IN (SELECT id FROM filtered_discussions)
),
note_rollup AS (
    SELECT
        discussion_id,
        SUM(CASE WHEN is_system = 0 THEN 1 ELSE 0 END) AS note_count,
        MAX(CASE WHEN rn = 1 AND is_system = 0 THEN author_username END) AS first_author,
        MAX(CASE WHEN rn = 1 AND is_system = 0 THEN body END) AS first_note_body,
        MAX(CASE WHEN position_new_path IS NOT NULL THEN position_new_path END) AS position_new_path,
        MAX(CASE WHEN position_new_line IS NOT NULL THEN position_new_line END) AS position_new_line
    FROM ranked_notes
    GROUP BY discussion_id
)
SELECT
    fd.id,
    fd.gitlab_discussion_id,
    fd.noteable_type,
    COALESCE(i.iid, m.iid) AS parent_iid,
    COALESCE(i.title, m.title) AS parent_title,
    p.path_with_namespace AS project_path,
    fd.individual_note,
    COALESCE(nr.note_count, 0) AS note_count,
    nr.first_author,
    nr.first_note_body,
    fd.first_note_at,
    fd.last_note_at,
    fd.resolvable,
    fd.resolved,
    nr.position_new_path,
    nr.position_new_line
FROM filtered_discussions fd
JOIN projects p ON fd.project_id = p.id
LEFT JOIN issues i ON fd.issue_id = i.id
LEFT JOIN merge_requests m ON fd.merge_request_id = m.id
LEFT JOIN note_rollup nr ON nr.discussion_id = fd.id
ORDER BY {sort_column} {order}
LIMIT ?
```

**Performance rationale**: The CTE pre-filters discussions before joining notes. The
`ranked_notes` CTE uses `ROW_NUMBER()` (a single pass over the notes index) instead of
correlated `(SELECT ... LIMIT 1)` sub-selects per discussion. For MR-scoped queries
(50-200 discussions) the performance is equivalent. For project-wide scans with thousands
of discussions, the window function approach avoids repeated index probes and produces a
more predictable query plan. The `MAX(CASE WHEN rn = 1 ...)` pattern extracts first-note
attributes from the grouped output without additional lookups.

**Note on SQLite FILTER syntax**: SQLite does not support `COUNT(*) FILTER (WHERE ...)`.
Use `SUM(CASE WHEN ... THEN 1 ELSE 0 END)` instead (as shown above).

#### 3c-ii. Note expansion query (--include-notes)

When `include_notes > 0`, after the main discussion query, run a follow-up query per
discussion to fetch its N most recent notes:

```sql
SELECT n.id, n.gitlab_id, n.author_username, n.body, n.note_type,
       n.is_system, n.created_at, n.updated_at,
       n.position_new_path, n.position_new_line,
       n.position_old_path, n.position_old_line,
       n.resolvable, n.resolved, n.resolved_by,
       d.noteable_type,
       COALESCE(i.iid, m.iid) AS parent_iid,
       COALESCE(i.title, m.title) AS parent_title,
       p.path_with_namespace AS project_path,
       d.gitlab_discussion_id
FROM notes n
JOIN discussions d ON n.discussion_id = d.id
JOIN projects p ON n.project_id = p.id
LEFT JOIN issues i ON d.issue_id = i.id
LEFT JOIN merge_requests m ON d.merge_request_id = m.id
WHERE d.id = ?
ORDER BY n.created_at DESC
LIMIT ?
```

**Optimization**: If discussion count is small (<= 50), batch all discussion IDs into a
single `WHERE d.id IN (?, ?, ...)` query with a secondary partition to split by discussion.
For larger result sets, fall back to per-discussion queries to avoid huge IN clauses. This
matches the existing note-loading pattern in `show.rs`.

The returned `NoteListRow` rows reuse the same struct and `NoteListRowJson` conversion from
workstream 1, ensuring identical note shape across all commands.

#### 3d. Filters struct

```rust
pub struct DiscussionListFilters {
    pub limit: usize,
    pub project: Option<String>,
    pub for_issue_iid: Option<i64>,
    pub for_mr_iid: Option<i64>,
    pub gitlab_discussion_id: Option<String>,
    pub resolution: Option<String>,
    pub since: Option<String>,
    pub path: Option<String>,
    pub noteable_type: Option<String>,
    pub sort: String,
    pub order: String,
    pub include_notes: usize,
}
```

Where-clause construction follows the exact pattern from `query_notes()`:
- `for_issue_iid` → subquery to resolve issue ID from IID + project
- `for_mr_iid` → subquery to resolve MR ID from IID + project
- `gitlab_discussion_id` → `d.gitlab_discussion_id = ?`
- `resolution` → `d.resolvable = 1 AND d.resolved = 0/1`
- `since` → `d.first_note_at >= ?` (using `parse_since()`)
- `path` → `EXISTS (SELECT 1 FROM notes n WHERE n.discussion_id = d.id AND n.position_new_path LIKE ?)`
- `noteable_type` → `d.noteable_type = ?`

#### 3e. Handler wiring

**File**: `src/main.rs`

Add match arm:

```rust
Some(Commands::Discussions(args)) => handle_discussions(cli.config.as_deref(), args, robot_mode),
```

Handler function:

```rust
fn handle_discussions(
    config_override: Option<&str>,
    args: DiscussionsArgs,
    robot_mode: bool,
) -> Result<(), Box<dyn std::error::Error>> {
    let start = std::time::Instant::now();
    let config = Config::load(config_override)?;
    let db_path = get_db_path(config.storage.db_path.as_deref());
    let conn = create_connection(&db_path)?;

    let order = if args.asc { "asc" } else { "desc" };
    let filters = DiscussionListFilters {
        limit: args.limit,
        project: args.project,
        for_issue_iid: args.for_issue,
        for_mr_iid: args.for_mr,
        gitlab_discussion_id: args.gitlab_discussion_id,
        resolution: args.resolution,
        since: args.since,
        path: args.path,
        noteable_type: args.noteable_type,
        sort: args.sort,
        order: order.to_string(),
        include_notes: args.include_notes,
    };

    let result = query_discussions(&conn, &filters, &config)?;

    let format = if robot_mode && args.format == "table" {
        "json"
    } else {
        &args.format
    };

    match format {
        "json" => print_list_discussions_json(
            &result,
            start.elapsed().as_millis() as u64,
            args.fields.as_deref(),
            robot_mode,
        ),
        "jsonl" => print_list_discussions_jsonl(&result),
        "csv" => print_list_discussions_csv(&result),
        _ => print_list_discussions(&result),
    }

    Ok(())
}
```

#### 3f. Print functions

**File**: `src/cli/commands/list.rs`

Follow same pattern as `print_list_notes_json`:

```rust
pub fn print_list_discussions_json(
    result: &DiscussionListResult,
    elapsed_ms: u64,
    fields: Option<&[String]>,
    robot_mode: bool,
) {
    let json_result = DiscussionListResultJson::from(result);
    let meta = RobotMeta { elapsed_ms };
    let output = serde_json::json!({
        "ok": true,
        "data": json_result,
        "meta": meta,
    });
    let mut output = output;
    if let Some(f) = fields {
        let expanded = expand_fields_preset(f, "discussions");
        if robot_mode {
            filter_fields_robot(&mut output, "discussions", &expanded);
        } else {
            filter_fields(&mut output, "discussions", &expanded);
        }
    }
    match serde_json::to_string(&output) {
        Ok(json) => println!("{json}"),
        Err(e) => eprintln!("Error serializing to JSON: {e}"),
    }
}
```

Table view: compact format showing discussion_id (first 8 chars), first author, note count,
resolved status, path, snippet.

CSV view: all fields, following same pattern as `print_list_notes_csv`.

#### 3g. Fields preset

**File**: `src/cli/robot.rs`

```rust
"discussions" => [
    "gitlab_discussion_id", "parent_iid", "note_count",
    "resolvable", "resolved", "first_author"
]
    .iter()
    .map(|s| (*s).to_string())
    .collect(),
```

### Tests

#### Test 1: Basic query returns discussions with gitlab_discussion_id

```rust
#[test]
fn query_discussions_basic() {
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_mr(&conn, 1, 1, 99, "Test MR");
    insert_discussion(&conn, 1, "hexhex123", 1, None, Some(1), "MergeRequest");
    insert_note_in_discussion(&conn, 1, 500, 1, 1, "alice", "first comment");
    insert_note_in_discussion(&conn, 2, 501, 1, 1, "bob", "reply");

    let filters = DiscussionListFilters::default_for_mr(99);
    let result = query_discussions(&conn, &filters, &Config::default()).unwrap();

    assert_eq!(result.discussions.len(), 1);
    assert_eq!(result.discussions[0].gitlab_discussion_id, "hexhex123");
    assert_eq!(result.discussions[0].note_count, 2);
    assert_eq!(result.discussions[0].first_author.as_deref(), Some("alice"));
}
```

#### Test 2: Resolution filter

```rust
#[test]
fn query_discussions_resolution_filter() {
    let conn = create_test_db();
    // Insert 2 discussions: one resolved, one unresolved
    // ...
    let filters = DiscussionListFilters {
        resolution: Some("unresolved".to_string()),
        ..DiscussionListFilters::default_for_mr(99)
    };
    let result = query_discussions(&conn, &filters, &Config::default()).unwrap();
    assert_eq!(result.total_count, 1);
    assert!(!result.discussions[0].resolved);
}
```

#### Test 3: Path filter

```rust
#[test]
fn query_discussions_path_filter() {
    // Insert discussions: one with diff notes on src/auth.rs, one general
    // Filter by path "src/auth.rs"
    // Assert only the diff note discussion is returned
}
```

#### Test 4: JSON serialization round-trip

```rust
#[test]
fn discussion_list_json_serialization() {
    let row = DiscussionListRow {
        id: 1,
        gitlab_discussion_id: "abc123".to_string(),
        noteable_type: "MergeRequest".to_string(),
        parent_iid: Some(99),
        parent_title: Some("Fix auth".to_string()),
        project_path: "group/repo".to_string(),
        individual_note: false,
        note_count: 3,
        first_author: Some("alice".to_string()),
        first_note_body: Some("This is a very long comment that should be truncated...".to_string()),
        first_note_at: 1_700_000_000_000,
        last_note_at: 1_700_001_000_000,
        resolvable: true,
        resolved: false,
        position_new_path: Some("src/auth.rs".to_string()),
        position_new_line: Some(42),
    };

    let json_row = DiscussionListRowJson::from(&row);
    let value = serde_json::to_value(&json_row).unwrap();
    assert_eq!(value["gitlab_discussion_id"], "abc123");
    assert_eq!(value["note_count"], 3);
    assert!(value["first_note_body_snippet"].as_str().unwrap().len() <= 120);
}
```

#### Test 5: Fields filtering works for discussions

```rust
#[test]
fn discussions_fields_minimal_preset() {
    let expanded = expand_fields_preset(&["minimal".to_string()], "discussions");
    assert!(expanded.contains(&"gitlab_discussion_id".to_string()));
    assert!(expanded.contains(&"parent_iid".to_string()));
}
```

#### Test 6: CTE query handles empty note_rollup gracefully

```rust
#[test]
fn query_discussions_with_no_notes() {
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_mr(&conn, 1, 1, 99, "Test MR");
    // Insert discussion with no notes (edge case: possible after sync issues)
    insert_discussion(&conn, 1, "orphan123", 1, None, Some(1), "MergeRequest");

    let filters = DiscussionListFilters::default_for_mr(99);
    let result = query_discussions(&conn, &filters, &Config::default()).unwrap();

    assert_eq!(result.discussions.len(), 1);
    assert_eq!(result.discussions[0].note_count, 0);
    assert!(result.discussions[0].first_author.is_none());
}
```

#### Test 7: --gitlab-discussion-id filter returns exact match

```rust
#[test]
fn query_discussions_by_gitlab_id() {
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_mr(&conn, 1, 1, 99, "Test MR");
    insert_discussion(&conn, 1, "target123", 1, None, Some(1), "MergeRequest");
    insert_discussion(&conn, 2, "other456", 1, None, Some(1), "MergeRequest");

    let filters = DiscussionListFilters {
        gitlab_discussion_id: Some("target123".to_string()),
        ..DiscussionListFilters::default_for_mr(99)
    };
    let result = query_discussions(&conn, &filters, &Config::default()).unwrap();

    assert_eq!(result.discussions.len(), 1);
    assert_eq!(result.discussions[0].gitlab_discussion_id, "target123");
}
```

#### Test 8: --include-notes populates notes array

```rust
#[test]
fn query_discussions_with_included_notes() {
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_mr(&conn, 1, 1, 99, "Test MR");
    insert_discussion(&conn, 1, "disc123", 1, None, Some(1), "MergeRequest");
    insert_note_in_discussion(&conn, 1, 500, 1, 1, "alice", "first");
    insert_note_in_discussion(&conn, 2, 501, 1, 1, "bob", "second");
    insert_note_in_discussion(&conn, 3, 502, 1, 1, "carol", "third");

    let filters = DiscussionListFilters {
        include_notes: 2,
        ..DiscussionListFilters::default_for_mr(99)
    };
    let result = query_discussions(&conn, &filters, &Config::default()).unwrap();

    assert_eq!(result.discussions.len(), 1);
    // Note: notes populated during JSON conversion, not in raw result
    // Test at handler/print level for full integration
}
```

#### Test 9: Bridge fields survive --fields filtering in robot mode

```rust
#[test]
fn discussions_bridge_fields_forced_in_robot_mode() {
    // Request only "note_count" — bridge fields must still appear
    let mut value = serde_json::json!({
        "data": {
            "discussions": [{
                "gitlab_discussion_id": "abc",
                "noteable_type": "MergeRequest",
                "parent_iid": 99,
                "project_path": "group/repo",
                "note_count": 3
            }]
        }
    });

    filter_fields_robot(
        &mut value,
        "discussions",
        &["note_count".to_string()],
    );

    let disc = &value["data"]["discussions"][0];
    assert_eq!(disc["note_count"], 3);
    assert!(disc.get("gitlab_discussion_id").is_some());
    assert!(disc.get("project_path").is_some());
}
```

---

## 4. Fix Robot-Docs Response Schemas

### Why

The notes command robot-docs says:
```json
"data": {"notes": "[NoteListRowJson]", "total_count": "int", "showing": "int"}
```

An agent sees `[NoteListRowJson]` — a Rust type name — and has no idea what fields are
available. Compare with the `issues` command which inline-lists every field. This forces
agents into trial-and-error field discovery.

### Changes Required

**File**: `src/main.rs`, robot-docs JSON block

#### 4a. Notes response_schema

Replace:
```json
"data": {"notes": "[NoteListRowJson]", "total_count": "int", "showing": "int"}
```

With:
```json
"data": {
    "notes": "[{id:int, gitlab_id:int, author_username:string, body:string?, note_type:string?, is_system:bool, created_at_iso:string, updated_at_iso:string, position_new_path:string?, position_new_line:int?, position_old_path:string?, position_old_line:int?, resolvable:bool, resolved:bool, resolved_by:string?, noteable_type:string?, parent_iid:int?, parent_title:string?, project_path:string, gitlab_discussion_id:string}]",
    "total_count": "int",
    "showing": "int"
}
```

#### 4b. Add discussions response_schema

```json
"discussions": {
    "description": "List discussions with thread-level metadata",
    "flags": [
        "--limit/-n <N>",
        "--for-issue <iid>",
        "--for-mr <iid>",
        "-p/--project <path>",
        "--gitlab-discussion-id <id>",
        "--resolution <unresolved|resolved>",
        "--since <period>",
        "--path <filepath>",
        "--noteable-type <Issue|MergeRequest>",
        "--include-notes <N>",
        "--sort <first_note|last_note>",
        "--asc",
        "--fields <list|minimal>",
        "--format <table|json|jsonl|csv>"
    ],
    "robot_flags": ["--format json", "--fields minimal"],
    "example": "lore --robot discussions --for-mr 99 --resolution unresolved",
    "response_schema": {
        "ok": "bool",
        "data": {
            "discussions": "[{gitlab_discussion_id:string, noteable_type:string, parent_iid:int?, parent_title:string?, project_path:string, individual_note:bool, note_count:int, first_author:string?, first_note_body_snippet:string?, first_note_at_iso:string, last_note_at_iso:string, resolvable:bool, resolved:bool, position_new_path:string?, position_new_line:int?, notes:[NoteListRowJson]?}]",
            "total_count": "int",
            "showing": "int"
        },
        "meta": {"elapsed_ms": "int"}
    }
}
```

#### 4c. Add to glab_equivalents

```json
{
    "glab": "glab api /projects/:id/merge_requests/:iid/discussions",
    "lore": "lore -J discussions --for-mr <iid>",
    "note": "Includes note counts, first author, resolution status, file positions"
}
```

#### 4d. Update show response_schema

Update the `issues` and `mrs` show schemas to reflect that `discussions` now include
`gitlab_discussion_id`, `resolvable`, `resolved`, and `last_note_at_iso`. Also reflect that
notes within show discussions now include `gitlab_note_id`.

#### 4e. Add to lore_exclusive list

```json
"discussions: Thread-level discussion listing with gitlab_discussion_id for API integration"
```

#### 4f. Add robot-docs contract tests

**File**: `src/main.rs` (within `#[cfg(test)]` module)

Add lightweight tests that parse the robot-docs JSON output and assert required Bridge
Contract fields are present. This prevents schema drift — if someone adds a field to the
struct but forgets to update robot-docs, the test fails.

```rust
#[test]
fn robot_docs_notes_schema_includes_bridge_fields() {
    let docs = get_robot_docs_json();  // helper that builds the robot-docs Value
    let notes_schema = docs["commands"]["notes"]["response_schema"]["data"]["notes"]
        .as_str().unwrap();
    assert!(notes_schema.contains("gitlab_discussion_id"));
    assert!(notes_schema.contains("project_path"));
    assert!(notes_schema.contains("parent_iid"));
}

#[test]
fn robot_docs_discussions_schema_includes_bridge_fields() {
    let docs = get_robot_docs_json();
    let disc_schema = docs["commands"]["discussions"]["response_schema"]["data"]["discussions"]
        .as_str().unwrap();
    assert!(disc_schema.contains("gitlab_discussion_id"));
    assert!(disc_schema.contains("project_path"));
    assert!(disc_schema.contains("parent_iid"));
}

#[test]
fn robot_docs_show_schema_includes_discussion_id() {
    let docs = get_robot_docs_json();
    // Verify issues and mrs show schemas reference gitlab_discussion_id
    // in their discussion sub-schemas
}
```

#### 4g. Add CLI-level contract integration tests

**File**: `src/cli/commands/list_tests.rs` or `src/main.rs` `#[cfg(test)]`

Add handler-level tests that invoke the command handlers with an in-memory DB and parse the
JSON output, asserting Bridge Contract fields are present. These are stronger than unit tests
on structs because they exercise the full path from query through serialization.

```rust
#[test]
fn notes_handler_json_includes_bridge_fields() {
    // Setup in-memory DB with project, discussion, note
    // Capture stdout from handle_notes (or call query_notes + print_list_notes_json)
    // Parse JSON, assert bridge fields present on every note
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_mr(&conn, 1, 1, 99, "Test MR");
    insert_discussion(&conn, 1, "abc123", 1, None, Some(1), "MergeRequest");
    insert_note_in_discussion(&conn, 1, 500, 1, 1, "alice", "hello");

    let result = query_notes(&conn, &NoteListFilters::default_for_mr(99), &Config::default()).unwrap();
    let json_result = NoteListResultJson::from(&result);
    let value = serde_json::to_value(&json_result).unwrap();

    for note in value["notes"].as_array().unwrap() {
        assert!(note.get("gitlab_discussion_id").is_some(), "missing gitlab_discussion_id");
        assert!(note.get("project_path").is_some(), "missing project_path");
        assert!(note.get("parent_iid").is_some(), "missing parent_iid");
    }
}

#[test]
fn discussions_handler_json_includes_bridge_fields() {
    let conn = create_test_db();
    insert_project(&conn, 1);
    insert_mr(&conn, 1, 1, 99, "Test MR");
    insert_discussion(&conn, 1, "abc123", 1, None, Some(1), "MergeRequest");
    insert_note_in_discussion(&conn, 1, 500, 1, 1, "alice", "hello");

    let result = query_discussions(&conn, &DiscussionListFilters::default_for_mr(99), &Config::default()).unwrap();
    let json_result = DiscussionListResultJson::from(&result);
    let value = serde_json::to_value(&json_result).unwrap();

    for disc in value["discussions"].as_array().unwrap() {
        assert!(disc.get("gitlab_discussion_id").is_some(), "missing gitlab_discussion_id");
        assert!(disc.get("project_path").is_some(), "missing project_path");
        assert!(disc.get("parent_iid").is_some(), "missing parent_iid");
    }
}
```

### Tests

Beyond the contract tests above, robot-docs changes are verified by running
`lore robot-docs` and inspecting output.

---

## Delivery Order

1. **Change 1** (notes output) — standalone, no dependencies. Can be released immediately.
2. **Change 2** (show output) — standalone, no dependencies. Can be released alongside 1.
3. **Change 3** (discussions command) — largest change, benefits from 1+2 being reviewed first
   to lock down field naming and serialization patterns.
4. **Change 4** (robot-docs + contract tests) — last, after all payloads are finalized.

Changes 1 and 2 can be done in parallel. Change 4 must come last since it documents the
final schema of all preceding changes.

**Cross-cutting**: The Bridge Contract field guardrail (force-including bridge fields in robot
mode) should be implemented as part of Change 1, since it modifies `filter_fields` in
`robot.rs` which all subsequent changes depend on. The `BRIDGE_FIELDS_*` constants are defined
once and reused by Changes 3 and 4.

---

## Validation Criteria

After all changes:

1. An agent can run `lore -J notes --for-mr 3929 --contains "really do prefer"` and get
   `gitlab_discussion_id` in the response
2. An agent can run `lore -J discussions --for-mr 3929 --resolution unresolved` to see all
   open threads with their IDs
3. An agent can run `lore -J mrs 3929` and see `gitlab_discussion_id`, `resolvable`,
   `resolved`, and `last_note_at_iso` on each discussion group, plus `gitlab_note_id` on
   each note within
4. `lore robot-docs` lists actual field names for all commands
5. All existing tests still pass
6. No clippy warnings (pedantic + nursery)
7. Robot-docs contract tests pass, preventing future schema drift
8. Bridge Contract fields (`project_path`, `noteable_type`, `parent_iid`,
   `gitlab_discussion_id`, `gitlab_note_id`) are present in every applicable read payload
9. Bridge Contract fields survive `--fields` filtering in robot mode (guardrail enforced)
10. `--gitlab-discussion-id` filter works on both `notes` and `discussions` commands
11. `--include-notes N` populates inline notes on `discussions` output
12. CLI-level contract integration tests verify bridge fields through the full handler path

---

## Rejected Recommendations

- **Rename `id`→`note_id` and `gitlab_id`→`gitlab_note_id` in notes list output** — rejected because every existing consumer (agents, scripts, field presets) uses `id` and `gitlab_id`. The fields are unambiguous within the `notes` context. The show-command note structs are a different story (they have no IDs at all), so we add `gitlab_note_id` there where it's genuinely missing. Renaming established fields is churn without proportional benefit.
- **Keyset cursor-based pagination (`--cursor` flag)** — rejected because no existing lore command has pagination, agents use `--limit` effectively, and adding a cursor mechanism is significant scope creep. Tracked as potential future work if agents hit real pagination needs.
- **Split `note_count` into `user_note_count`/`total_note_count` and rename `first_author` to `first_user_author`** — rejected because `note_count` already excludes system notes by query design (the `WHERE is_system = 0` / `CASE WHEN` filter), and `first_author` already targets the first non-system note. The current naming is clear and consistent with how `notes --include-system` works elsewhere.
- **Match path filter on both `position_new_path` and `position_old_path`** — rejected because agents care about where code is *now* (new path), not where it was before a rename. Matching old paths adds complexity and returns confusing results for moved files.
- **Separate migration file for discussion-list indexes** — rejected because this project uses a `MIGRATIONS` array in `src/core/db.rs`, not separate migration files. If profiling shows the new query needs indexes, they'll be added to the migration array in the standard way. Premature index creation without measurement is against project practice.
- **Shared contract model / workstream 0 (shared constants module)** — rejected because 4 structs sharing field names in a codebase this size isn't drift-prone. We have compile-time contract tests (robot-docs assertions + handler-level JSON tests) that catch drift. A constants module for field name strings adds indirection without proportional gain. The Bridge Contract field guardrail (`BRIDGE_FIELDS_*` arrays in robot.rs) provides the centralized definition where it matters — at the filtering enforcement point.
- **Structured robot-docs schema (JSON objects instead of string blobs)** — rejected because the current compact string format is intentionally token-efficient for agent consumption. Switching to nested JSON objects per field would significantly bloat robot-docs output. The string-based contract tests are sufficient — they test what agents actually parse. Agents already work with the inline field listing format used by `issues` and `mrs`.
- **`bridge_contract` meta-section in robot-docs output** — rejected because agents don't need a separate meta-contract section; they need correct field listings per command, which we already provide. Adding a cross-cutting contract section to robot-docs adds documentation surface area without improving the agent workflow.
- **Performance regression benchmark test (ignored by default)** — rejected because timing-based assertions are inherently flaky across machines, CI environments, and load conditions. Performance is validated through query plan analysis (EXPLAIN) and manual profiling, not hard-coded elapsed-time thresholds.