From 93c7b3d6a3ea1c7eb24d54ab7730fabbdf8bddc9 Mon Sep 17 00:00:00 2001
From: YellowSnnowmann <167776381+YellowSnnowmann@users.noreply.github.com>
Date: Tue, 16 Jun 2026 05:50:08 +0530
Subject: [PATCH] fix(agent): parse Claude-native <invoke name> attribute tool
 calls (#3622)

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Steven Enamakel <enamakel@tinyhumans.ai>
---
 .../agent/harness/harness_gap_tests.rs        | 108 +++++++++++++++
 src/openhuman/agent/harness/parse.rs          | 129 +++++++++++++++++-
 2 files changed, 236 insertions(+), 1 deletion(-)

diff --git a/src/openhuman/agent/harness/harness_gap_tests.rs b/src/openhuman/agent/harness/harness_gap_tests.rs
index c6007e364..afb069b60 100644
--- a/src/openhuman/agent/harness/harness_gap_tests.rs
+++ b/src/openhuman/agent/harness/harness_gap_tests.rs
@@ -475,6 +475,114 @@ fn parse_tool_calls_invoke_tag_with_json_body() {
     );
 }
 
+// ─────────────────────────────────────────────────────────────────────────────
+// Item 3b — parse_tool_calls: Claude-native <invoke name="…"> attribute form
+//           with nested <parameter name="…"> children (issue #3493).
+//           Claude-family models ignore the injected <tool_call>{json} template
+//           and emit their trained syntax; the parser must recover it instead of
+//           leaking the raw markup as assistant text.
+// ─────────────────────────────────────────────────────────────────────────────
+
+#[test]
+fn parse_tool_calls_invoke_attribute_form_single_param() {
+    use crate::openhuman::agent::harness::parse::parse_tool_calls;
+
+    let input =
+        "Sure.\n<invoke name=\"echo\">\n<parameter name=\"value\">hi</parameter>\n</invoke>\ndone";
+    let (text, calls) = parse_tool_calls(input);
+
+    assert_eq!(
+        calls.len(),
+        1,
+        "attribute-form <invoke> should parse one call"
+    );
+    assert_eq!(calls[0].name, "echo");
+    assert_eq!(calls[0].arguments, serde_json::json!({"value": "hi"}));
+    // Surrounding prose preserved; raw <invoke> markup must not leak.
+    assert!(text.contains("Sure."), "text before tag preserved");
+    assert!(text.contains("done"), "text after tag preserved");
+    assert!(
+        !text.contains("<invoke"),
+        "raw <invoke> markup must not surface in assistant text"
+    );
+    assert!(
+        !text.contains("<parameter"),
+        "raw <parameter> markup must not surface in assistant text"
+    );
+}
+
+#[test]
+fn parse_tool_calls_invoke_attribute_form_multiple_params_scalar_policy() {
+    use crate::openhuman::agent::harness::parse::parse_tool_calls;
+
+    // Multiple <parameter> children. Scalar policy: a value that parses as JSON
+    // (number, bool) becomes that JSON type; anything else stays a string. A
+    // parameter with an empty name is skipped (it cannot key an argument).
+    let input = concat!(
+        "<invoke name=\"search\">\n",
+        "<parameter name=\"query\">rust parsers</parameter>\n",
+        "<parameter name=\"limit\">5</parameter>\n",
+        "<parameter name=\"fuzzy\">true</parameter>\n",
+        "<parameter name=\"\">ignored</parameter>\n",
+        "</invoke>"
+    );
+    let (_text, calls) = parse_tool_calls(input);
+
+    assert_eq!(calls.len(), 1, "should parse one call");
+    assert_eq!(calls[0].name, "search");
+    assert_eq!(
+        calls[0].arguments,
+        serde_json::json!({"query": "rust parsers", "limit": 5, "fuzzy": true})
+    );
+}
+
+#[test]
+fn parse_tool_calls_invoke_attribute_form_missing_close_tag_is_text() {
+    use crate::openhuman::agent::harness::parse::parse_tool_calls;
+
+    // No closing </invoke>: nothing to dispatch. The block is left as text
+    // rather than silently dropped.
+    let input = "before\n<invoke name=\"echo\">\n<parameter name=\"v\">hi</parameter>";
+    let (text, calls) = parse_tool_calls(input);
+
+    assert_eq!(calls.len(), 0, "unterminated <invoke> yields no calls");
+    assert!(text.contains("before"), "preceding text preserved");
+    assert!(
+        text.contains("<invoke"),
+        "unterminated block left as text, not dropped"
+    );
+}
+
+#[test]
+fn parse_tool_calls_invoke_attribute_form_missing_name_is_text() {
+    use crate::openhuman::agent::harness::parse::parse_tool_calls;
+
+    // Attribute form without a `name` attribute cannot name a tool → no call.
+    let input = "<invoke foo=\"bar\">\n<parameter name=\"v\">hi</parameter>\n</invoke>";
+    let (_text, calls) = parse_tool_calls(input);
+
+    assert_eq!(calls.len(), 0, "missing name attribute yields no calls");
+}
+
+#[test]
+fn parse_tool_calls_mixed_tool_call_json_and_invoke_attribute() {
+    use crate::openhuman::agent::harness::parse::parse_tool_calls;
+
+    // A canonical <tool_call>{json} block and a Claude-native attribute-form
+    // <invoke> block in the same response are both recovered, earliest first.
+    let input = concat!(
+        "<tool_call>{\"name\":\"first\",\"arguments\":{\"a\":1}}</tool_call>\n",
+        "<invoke name=\"second\">\n<parameter name=\"b\">two</parameter>\n</invoke>"
+    );
+    let (_text, calls) = parse_tool_calls(input);
+
+    assert_eq!(calls.len(), 2, "both tag forms parsed");
+    assert_eq!(calls[0].name, "first");
+    assert_eq!(calls[0].arguments, serde_json::json!({"a": 1}));
+    assert_eq!(calls[1].name, "second");
+    assert_eq!(calls[1].arguments, serde_json::json!({"b": "two"}));
+}
+
 #[test]
 fn parse_tool_calls_markdown_fence_yaml_like_json_body() {
     use crate::openhuman::agent::harness::parse::parse_tool_calls;
diff --git a/src/openhuman/agent/harness/parse.rs b/src/openhuman/agent/harness/parse.rs
index b1265c267..f8863b011 100644
--- a/src/openhuman/agent/harness/parse.rs
+++ b/src/openhuman/agent/harness/parse.rs
@@ -183,6 +183,92 @@ pub(crate) fn matching_tool_call_close_tag(open_tag: &str) -> Option<&'static st
     }
 }
 
+/// `<invoke` prefix shared by the bare (`<invoke>`) and Claude-native
+/// attribute (`<invoke name="…">`) forms.
+const INVOKE_PREFIX: &str = "<invoke";
+
+/// Locate the earliest Claude-native attribute-form `<invoke …>` open tag
+/// (issue #3493). Matches `<invoke` only when the next character is whitespace
+/// — i.e. attributes follow. The bare `<invoke>` form (next char `>`) is
+/// intentionally skipped here; it is recognised as a literal tag with a JSON
+/// body via [`TOOL_CALL_OPEN_TAGS`], preserving back-compat.
+fn find_invoke_attr_tag(haystack: &str) -> Option<usize> {
+    let mut from = 0;
+    while let Some(rel) = haystack[from..].find(INVOKE_PREFIX) {
+        let idx = from + rel;
+        let after = &haystack[idx + INVOKE_PREFIX.len()..];
+        match after.chars().next() {
+            Some(c) if c.is_whitespace() => return Some(idx),
+            _ => from = idx + INVOKE_PREFIX.len(),
+        }
+    }
+    None
+}
+
+/// Scalar policy for `<parameter>` values: a value that parses as JSON
+/// (number, bool, null, array, object) is kept as that JSON type; anything
+/// else — the common case of bare text — stays a string. Mirrors the tolerant
+/// arg handling in [`parse_arguments_value`].
+fn parameter_scalar_value(raw: &str) -> serde_json::Value {
+    let trimmed = raw.trim();
+    match serde_json::from_str::<serde_json::Value>(trimmed) {
+        Ok(
+            value @ (serde_json::Value::Number(_)
+            | serde_json::Value::Bool(_)
+            | serde_json::Value::Null
+            | serde_json::Value::Array(_)
+            | serde_json::Value::Object(_)),
+        ) => value,
+        _ => serde_json::Value::String(trimmed.to_string()),
+    }
+}
+
+/// Parse a Claude-native attribute-form invoke block whose text begins
+/// immediately after the `<invoke` prefix (at the attributes). Returns the
+/// recovered call and the number of bytes consumed up to and including the
+/// closing `</invoke>`. `None` when the `name` attribute or the closing tag is
+/// missing — the caller then leaves the markup as text rather than dropping it.
+fn parse_invoke_attribute_block(after_prefix: &str) -> Option<(ParsedToolCall, usize)> {
+    static INVOKE_NAME_RE: LazyLock<Regex> =
+        LazyLock::new(|| Regex::new(r#"name\s*=\s*"([^"]*)""#).unwrap());
+    static PARAMETER_RE: LazyLock<Regex> = LazyLock::new(|| {
+        Regex::new(r#"(?s)<parameter\s+name\s*=\s*"([^"]*)"\s*>(.*?)</parameter>"#).unwrap()
+    });
+
+    let open_end = after_prefix.find('>')?;
+    let attrs = &after_prefix[..open_end];
+    let name = INVOKE_NAME_RE
+        .captures(attrs)
+        .and_then(|c| c.get(1))
+        .map(|m| m.as_str().trim().to_string())
+        .filter(|n| !n.is_empty())?;
+
+    let body = &after_prefix[open_end + 1..];
+    let close_rel = body.find("</invoke>")?;
+    let inner = &body[..close_rel];
+
+    let mut arguments = serde_json::Map::new();
+    for cap in PARAMETER_RE.captures_iter(inner) {
+        // Groups 1 (name) and 2 (value) are mandatory in the pattern, so a
+        // captured match always has both — index access is safe.
+        let key = cap[1].trim();
+        if key.is_empty() {
+            continue;
+        }
+        arguments.insert(key.to_string(), parameter_scalar_value(&cap[2]));
+    }
+
+    let consumed = open_end + 1 + close_rel + "</invoke>".len();
+    Some((
+        ParsedToolCall {
+            name,
+            arguments: serde_json::Value::Object(arguments),
+            id: None,
+        },
+        consumed,
+    ))
+}
+
 pub(crate) fn extract_first_json_value_with_end(input: &str) -> Option<(serde_json::Value, usize)> {
     let trimmed = input.trim_start();
     let trim_offset = input.len().saturating_sub(trimmed.len());
@@ -439,7 +525,48 @@ pub(crate) fn parse_tool_calls(response: &str) -> (String, Vec<ParsedToolCall>)
     }
 
     // Fall back to XML-style tool-call tag parsing.
-    while let Some((start, open_tag)) = find_first_tag(remaining, &TOOL_CALL_OPEN_TAGS) {
+    loop {
+        let literal = find_first_tag(remaining, &TOOL_CALL_OPEN_TAGS);
+        let invoke_attr = find_invoke_attr_tag(remaining);
+
+        // Choose the earliest-positioned recognised open tag. The bare
+        // `<invoke>` literal and the attribute form `<invoke …>` never collide
+        // at one offset (one is followed by `>`, the other by whitespace), so a
+        // simple index comparison disambiguates them (issue #3493).
+        let use_invoke_attr = match (invoke_attr, literal.as_ref()) {
+            (Some(i), Some((l, _))) => i < *l,
+            (Some(_), None) => true,
+            _ => false,
+        };
+
+        if use_invoke_attr {
+            let start = invoke_attr.expect("use_invoke_attr implies Some");
+            let before = &remaining[..start];
+            if !before.trim().is_empty() {
+                text_parts.push(before.trim().to_string());
+            }
+
+            let after_prefix = &remaining[start + INVOKE_PREFIX.len()..];
+            if let Some((parsed, consumed)) = parse_invoke_attribute_block(after_prefix) {
+                calls.push(parsed);
+                remaining = &after_prefix[consumed..];
+                continue;
+            }
+
+            // Unparseable attribute-form block (no `name`/no close tag): leave
+            // it and the rest as text instead of silently dropping content.
+            tracing::warn!(
+                body_chars = after_prefix.chars().count(),
+                "[agent_parse] malformed <invoke> attribute block: missing name or close tag"
+            );
+            remaining = &remaining[start..];
+            break;
+        }
+
+        let Some((start, open_tag)) = literal else {
+            break;
+        };
+
         // Everything before the tag is text.
         let before = &remaining[..start];
         if !before.trim().is_empty() {