feat(cli): surface stored-query @description/@instruction in queries list (#280)

* test: e2e coverage for @description/@instruction surfaces

Add end-to-end tests pinning the two annotation surfaces as they exist
today, at their real boundaries:

- engine (lifecycle.rs): schema-level @description (node/edge/property)
  and @instruction (node/edge) persist verbatim into the on-disk
  _schema.ir.json through Omnigraph::init; property-level @instruction
  aborts init and writes no schema IR.
- server (stored_queries.rs): query-level @description/@instruction on a
  stored query surface as typed QueryCatalogEntry fields over
  GET /queries, and a query declaring neither omits both fields.

No behavior change — these document the current contract.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(cli): surface stored-query @description/@instruction in `queries list`

A stored query's @description/@instruction are its catalog metadata —
what it does and how to invoke it. The HTTP GET /queries catalog already
carries them, but `omnigraph queries list` dropped both fields in human
and --json output even though they were available on the registry entry.

Carry description/instruction on QueriesListItem (Option, skipped when
None) and copy them from the query decl. Human output prints an indented
`description:` / `instruction:` line per query when present; --json
includes the fields when present and omits them otherwise — matching the
HTTP catalog shape documented in docs/user/operations/server.md.

Tests (cli_queries.rs): a query with both annotations surfaces them in
human + --json; a query with neither prints no annotation lines and omits
both JSON fields.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(cli): document `queries list` output incl. description/instruction

Per AGENTS.md maintenance Rule 1, document the user-visible `queries list`
output alongside the field addition. The `queries` command family had no
row in the CLI reference top-level table; add one covering `list` (human
+ --json shapes, with description/instruction shown only when declared,
matching the HTTP GET /queries catalog) and `validate`.

Addresses the Greptile P2 review finding on PR #280.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(cli): indent multiline stored-query annotations in `queries list`

A `@description`/`@instruction` value can be multiline (GQ string literals
admit newlines), which made the human `queries list` output break back to
the left margin on continuation lines. Indent continuation lines to align
under the first via a `print_query_annotation` helper. Addresses review
feedback from @martin-g on PR #280.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Ragnor Comerford <ragnor.comerford@gmail.com>

crates/omnigraph-cli/src/helpers.rs

@@ -875,6 +875,25 @@ pub(crate) async fn execute_queries_validate(
     Ok(())
 }
 
+/// Print a stored-query annotation under its `queries list` entry. A
+/// `@description`/`@instruction` value may be multiline (GQ string literals
+/// admit newlines); continuation lines are indented to align under the first
+/// so the catalog stays readable instead of breaking the left margin.
+fn print_query_annotation(label: &str, value: &str) {
+    let prefix = format!("    {label}: ");
+    let continuation = " ".repeat(prefix.len());
+    let mut lines = value.split('\n');
+    match lines.next() {
+        Some(first) => {
+            println!("{prefix}{first}");
+            for line in lines {
+                println!("{continuation}{line}");
+            }
+        }
+        None => println!("{prefix}"),
+    }
+}
+
 /// `queries list --cluster <dir>` (RFC-011): list the catalog's stored queries.
 /// With `--graph`, scope to one graph.
 pub(crate) async fn execute_queries_list(
@@ -893,6 +912,8 @@ pub(crate) async fn execute_queries_list(
                 mcp_expose: q.expose,
                 tool_name: q.tool_name.clone(),
                 mutation: q.is_mutation(),
+                description: q.decl.description.clone(),
+                instruction: q.decl.instruction.clone(),
                 params: q
                     .decl
                     .params
@@ -933,6 +954,12 @@ pub(crate) async fn execute_queries_list(
                 String::new()
             };
             println!("{kind}  {}({params}){mcp}", q.name);
+            if let Some(description) = &q.description {
+                print_query_annotation("description", description);
+            }
+            if let Some(instruction) = &q.instruction {
+                print_query_annotation("instruction", instruction);
+            }
         }
     }
     Ok(())

crates/omnigraph-cli/src/output.rs

@@ -849,6 +849,13 @@ pub(crate) struct QueriesListItem {
     pub(crate) mcp_expose: bool,
     pub(crate) tool_name: Option<String>,
     pub(crate) mutation: bool,
+    /// `@description` from the query declaration — what the query is for.
+    /// Carried so the CLI catalog matches the HTTP `GET /queries` surface.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub(crate) description: Option<String>,
+    /// `@instruction` from the query declaration — how/when to invoke it.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub(crate) instruction: Option<String>,
     pub(crate) params: Vec<QueriesParam>,
 }
 

crates/omnigraph-cli/tests/cli_queries.rs

@@ -231,6 +231,125 @@ fn queries_list_prints_registered_query() {
     );
 }
 
+#[test]
+fn queries_list_surfaces_description_and_instruction() {
+    // `@description`/`@instruction` are the whole point of a stored query in a
+    // catalog — they tell an agent/operator what it does and how to invoke it.
+    // The CLI catalog must surface them in both human and --json output, to
+    // match the HTTP `GET /queries` surface.
+    let cluster = converged_cluster_with_query(
+        "described.gq",
+        "query described($name: String) \
+            @description(\"Find a person by exact name.\") \
+            @instruction(\"Use for exact lookups; prefer search for fuzzy matches.\") \
+            { match { $p: Person { name: $name } } return { $p.age } }",
+        "      described:\n        file: ./described.gq\n",
+    );
+
+    // Human output.
+    let output = output_success(
+        cli().arg("queries").arg("list").arg("--cluster").arg(cluster.path()),
+    );
+    let stdout = stdout_string(&output);
+    assert!(
+        stdout.contains("description: Find a person by exact name."),
+        "human list must show @description; stdout:\n{stdout}"
+    );
+    assert!(
+        stdout.contains("instruction: Use for exact lookups; prefer search for fuzzy matches."),
+        "human list must show @instruction; stdout:\n{stdout}"
+    );
+
+    // --json output.
+    let output = output_success(
+        cli()
+            .arg("queries")
+            .arg("list")
+            .arg("--cluster")
+            .arg(cluster.path())
+            .arg("--json"),
+    );
+    let body: serde_json::Value = serde_json::from_slice(&output.stdout).unwrap();
+    let entry = body["queries"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .find(|q| q["name"] == "described")
+        .unwrap();
+    assert_eq!(entry["description"], "Find a person by exact name.");
+    assert_eq!(
+        entry["instruction"],
+        "Use for exact lookups; prefer search for fuzzy matches."
+    );
+}
+
+#[test]
+fn queries_list_indents_multiline_annotation_continuation() {
+    // GQ string literals admit newlines, so a `@description`/`@instruction`
+    // can be multiline. Human output must indent continuation lines to align
+    // under the first rather than breaking back to the left margin.
+    let cluster = converged_cluster_with_query(
+        "multi.gq",
+        "query multi($name: String) \
+            @description(\"line one\\nline two\") \
+            { match { $p: Person { name: $name } } return { $p.age } }",
+        "      multi:\n        file: ./multi.gq\n",
+    );
+    let output = output_success(
+        cli().arg("queries").arg("list").arg("--cluster").arg(cluster.path()),
+    );
+    let stdout = stdout_string(&output);
+    // "    description: " is 17 chars wide; the continuation aligns under it.
+    assert!(
+        stdout.contains("    description: line one\n                 line two"),
+        "multiline annotation must indent the continuation; stdout:\n{stdout}"
+    );
+}
+
+#[test]
+fn queries_list_omits_annotations_when_absent() {
+    // The other half of the contract: a query that declares neither annotation
+    // prints no extra lines and omits both JSON fields entirely. This keeps the
+    // catalog clean rather than echoing empty `description:`/`instruction:`.
+    let cluster = converged_cluster_with_query(
+        "bare.gq",
+        "query bare() { match { $p: Person } return { $p.name } }",
+        "      bare:\n        file: ./bare.gq\n",
+    );
+
+    // Human output: the query is listed, but no annotation lines.
+    let output = output_success(
+        cli().arg("queries").arg("list").arg("--cluster").arg(cluster.path()),
+    );
+    let stdout = stdout_string(&output);
+    assert!(stdout.contains("bare()"), "stdout:\n{stdout}");
+    assert!(
+        !stdout.contains("description:") && !stdout.contains("instruction:"),
+        "a query without annotations prints no annotation lines; stdout:\n{stdout}"
+    );
+
+    // --json output: both fields omitted (not present as null).
+    let output = output_success(
+        cli()
+            .arg("queries")
+            .arg("list")
+            .arg("--cluster")
+            .arg(cluster.path())
+            .arg("--json"),
+    );
+    let body: serde_json::Value = serde_json::from_slice(&output.stdout).unwrap();
+    let entry = body["queries"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .find(|q| q["name"] == "bare")
+        .unwrap();
+    assert!(
+        entry.get("description").is_none() && entry.get("instruction").is_none(),
+        "a query without annotations omits both JSON fields: {entry}"
+    );
+}
+
 #[test]
 fn queries_validate_requires_a_cluster() {
     // RFC-011: with no --cluster (and no cluster profile), the command errors

crates/omnigraph-server/tests/stored_queries.rs

@@ -369,6 +369,47 @@ async fn list_queries_is_read_gated_so_a_non_invoker_can_list() {
     );
 }
 
+#[tokio::test(flavor = "multi_thread")]
+async fn list_queries_surfaces_query_description_and_instruction() {
+    // E2e for the query-level `.gq` surface: `@description`/`@instruction` on
+    // a stored query declaration are carried through to clients via the typed
+    // `QueryCatalogEntry` fields over `GET /queries`. A query without them
+    // omits both fields (serde `skip_serializing_if = "Option::is_none"`).
+    let described = "query described($name: String) \
+        @description(\"Find a person by exact name.\") \
+        @instruction(\"Use for exact lookups; prefer search for fuzzy matches.\") \
+        { match { $p: Person { name: $name } } return { $p.age } }";
+    let (_temp, app) = app_with_stored_queries(
+        &[
+            ("described", described, true),
+            ("bare", "query bare() { match { $p: Person } return { $p.name } }", true),
+        ],
+        &[("act-invoke", "t-invoke")],
+        INVOKE_POLICY_YAML,
+    )
+    .await;
+    let (status, body) = json_response(&app, get_request(&g("/queries"), "t-invoke")).await;
+    assert_eq!(status, StatusCode::OK, "body: {body}");
+    let entries = body["queries"].as_array().unwrap();
+
+    let described = entries.iter().find(|q| q["name"] == "described").unwrap();
+    assert_eq!(
+        described["description"], "Find a person by exact name.",
+        "query @description surfaces over GET /queries: {described}"
+    );
+    assert_eq!(
+        described["instruction"],
+        "Use for exact lookups; prefer search for fuzzy matches.",
+        "query @instruction surfaces over GET /queries: {described}"
+    );
+
+    let bare = entries.iter().find(|q| q["name"] == "bare").unwrap();
+    assert!(
+        bare.get("description").is_none() && bare.get("instruction").is_none(),
+        "a query without the annotations omits both fields: {bare}"
+    );
+}
+
 #[tokio::test(flavor = "multi_thread")]
 async fn list_queries_is_empty_when_no_registry() {
     let (_temp, app) = app_for_loaded_graph_with_auth("demo-token").await;

crates/omnigraph/tests/lifecycle.rs

@@ -304,3 +304,108 @@ async fn init_with_force_recovers_from_orphan_schema_files() {
         "force-recovered graph must have full schema state written"
     );
 }
+
+/// E2e for the schema-level `.pg` surface: `@description` (node / edge /
+/// property) and `@instruction` (node / edge only) parse, validate, and
+/// persist verbatim into the on-disk `_schema.ir.json` through `Omnigraph::init`
+/// — the contract that surfaces them in catalog metadata for tooling.
+#[tokio::test]
+async fn schema_annotations_persist_into_ir_json_on_init() {
+    let dir = tempfile::tempdir().unwrap();
+    let uri = dir.path().to_str().unwrap();
+
+    let schema = r#"
+node Task @description("Tracked work item") @instruction("Prefer querying by slug") {
+    slug: String @key @description("Stable external identifier")
+}
+
+edge DependsOn: Task -> Task @description("Hard dependency") @instruction("Use only for blockers")
+"#;
+
+    Omnigraph::init(uri, schema).await.unwrap();
+
+    let ir_json = fs::read_to_string(dir.path().join("_schema.ir.json")).unwrap();
+    let ir: serde_json::Value = serde_json::from_str(&ir_json).unwrap();
+
+    // Helper: collect the {name -> value} map of annotations that carry a
+    // string value. Value-less annotations (e.g. `@key`, which also desugars
+    // to a constraint) are skipped — they aren't what this test asserts.
+    let anns = |v: &serde_json::Value| -> std::collections::BTreeMap<String, String> {
+        v["annotations"]
+            .as_array()
+            .unwrap()
+            .iter()
+            .filter_map(|a| {
+                Some((
+                    a["name"].as_str()?.to_string(),
+                    a["value"].as_str()?.to_string(),
+                ))
+            })
+            .collect()
+    };
+
+    let node = ir["nodes"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .find(|n| n["name"] == "Task")
+        .unwrap();
+    let node_anns = anns(node);
+    assert_eq!(node_anns.get("description").map(String::as_str), Some("Tracked work item"));
+    assert_eq!(
+        node_anns.get("instruction").map(String::as_str),
+        Some("Prefer querying by slug"),
+        "node @instruction persists into _schema.ir.json"
+    );
+
+    let prop = node["properties"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .find(|p| p["name"] == "slug")
+        .unwrap();
+    assert_eq!(
+        anns(prop).get("description").map(String::as_str),
+        Some("Stable external identifier"),
+        "property @description persists into _schema.ir.json"
+    );
+
+    let edge = ir["edges"]
+        .as_array()
+        .unwrap()
+        .iter()
+        .find(|e| e["name"] == "DependsOn")
+        .unwrap();
+    let edge_anns = anns(edge);
+    assert_eq!(edge_anns.get("description").map(String::as_str), Some("Hard dependency"));
+    assert_eq!(edge_anns.get("instruction").map(String::as_str), Some("Use only for blockers"));
+}
+
+/// `@instruction` is rejected on a property at compile time, so init aborts
+/// before any graph state is written (mirrors the parser-level rejection from
+/// the full engine boundary).
+#[tokio::test]
+async fn init_rejects_instruction_on_property() {
+    let dir = tempfile::tempdir().unwrap();
+    let uri = dir.path().to_str().unwrap();
+
+    let schema = r#"
+node Task {
+    slug: String @key @instruction("bad")
+}
+"#;
+
+    // `Omnigraph` is not `Debug`, so match rather than `unwrap_err`.
+    let err = match Omnigraph::init(uri, schema).await {
+        Ok(_) => panic!("property-level @instruction must abort init"),
+        Err(err) => err,
+    };
+    assert!(
+        err.to_string().contains("@instruction is only supported on node and edge types"),
+        "property-level @instruction must abort init: {err}"
+    );
+    assert!(
+        !dir.path().join("_schema.ir.json").exists(),
+        "rejected init must not persist a schema IR"
+    );
+}