set_status: reject multi-line and over-200-char text (#720)

hive-c0re/src/limits.rs

@@ -41,6 +41,56 @@ pub fn check_size(label: &str, body: &str) -> Result<(), String> {
     }
 }
 
+/// Per-status soft cap. `set_status` renders as a short chip on the
+/// dashboard agent card — the front-end truncates long strings to
+/// keep the row layout intact, so a multi-paragraph "session report"
+/// is wasted bytes that just bloat the rescan emit + container view
+/// payload. Cap at 200 chars to fit the chip plus a little
+/// descriptive padding without forcing the operator to read a
+/// scrolling chunk.
+pub const STATUS_MAX_CHARS: usize = 200;
+
+/// Validate a `set_status` payload (#720). Single-line + bounded so
+/// callers can't dump multi-paragraph session reports into the
+/// dashboard chip. Whitespace trim is done by the caller before the
+/// store-to-disk step — we run validation on the trimmed form so
+/// surrounding whitespace doesn't push a borderline-legal status
+/// past the cap.
+///
+/// Empty / all-whitespace input is accepted: the call site treats
+/// that as "clear the status" and removes the on-disk sentinel. Tests
+/// + caller cover both directions.
+///
+/// Returns a caller-ready error string suitable for surfacing in the
+/// `*Response::Err` shape.
+pub fn check_status_text(text: &str) -> Result<(), String> {
+    let trimmed = text.trim();
+    if trimmed.is_empty() {
+        // Empty = clear-status sentinel; nothing to validate.
+        return Ok(());
+    }
+    // Newline / carriage-return: status is a single-line chip on the
+    // dashboard. A multi-line argus session report is the canonical
+    // failure mode from #720.
+    if trimmed.contains('\n') || trimmed.contains('\r') {
+        return Err(
+            "set_status text must be a single line — write multi-line context to \
+             a file under your `/agents/<you>/state/` dir and reference that path \
+             from the chip instead"
+                .to_owned(),
+        );
+    }
+    let len = trimmed.chars().count();
+    if len > STATUS_MAX_CHARS {
+        return Err(format!(
+            "set_status text too long ({len} chars, max {STATUS_MAX_CHARS}); the \
+             dashboard chip truncates anything longer, so trim to a short summary \
+             and write the detail to `/agents/<you>/state/<file>` instead"
+        ));
+    }
+    Ok(())
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -63,4 +113,62 @@ mod tests {
         let err = check_size("question", &"x".repeat(MESSAGE_MAX_BYTES + 1)).unwrap_err();
         assert!(err.starts_with("question body too long"));
     }
+
+    #[test]
+    fn check_status_accepts_short_single_line() {
+        assert!(check_status_text("idle").is_ok());
+        assert!(check_status_text("processing matrix messages").is_ok());
+        // Boundary: exactly STATUS_MAX_CHARS chars trimmed is still
+        // accepted; one more rejects.
+        let max = "a".repeat(STATUS_MAX_CHARS);
+        assert!(check_status_text(&max).is_ok());
+    }
+
+    #[test]
+    fn check_status_accepts_empty_and_whitespace() {
+        // Empty + whitespace-only are the "clear status" sentinel and
+        // bypass the rest of the checks.
+        assert!(check_status_text("").is_ok());
+        assert!(check_status_text("   ").is_ok());
+        assert!(check_status_text("\n\t  ").is_ok());
+    }
+
+    #[test]
+    fn check_status_rejects_multi_line() {
+        let err = check_status_text("line one\nline two").unwrap_err();
+        assert!(err.contains("single line"), "err = {err}");
+        // Carriage return alone also rejects (windows linebreak / CR-only).
+        assert!(check_status_text("a\rb").is_err());
+    }
+
+    #[test]
+    fn check_status_rejects_oversize() {
+        let too_long = "a".repeat(STATUS_MAX_CHARS + 1);
+        let err = check_status_text(&too_long).unwrap_err();
+        assert!(err.contains("too long"), "err = {err}");
+        assert!(err.contains(&format!("max {STATUS_MAX_CHARS}")));
+    }
+
+    #[test]
+    fn check_status_counts_chars_not_bytes() {
+        // Multi-byte chars (emoji, accented letters) count once each
+        // per char — chars().count() not byte len. STATUS_MAX_CHARS
+        // worth of 4-byte chars is still legal.
+        let emoji = "💜".repeat(STATUS_MAX_CHARS);
+        assert!(
+            check_status_text(&emoji).is_ok(),
+            "{STATUS_MAX_CHARS} emoji should fit"
+        );
+        let too_many = "💜".repeat(STATUS_MAX_CHARS + 1);
+        assert!(check_status_text(&too_many).is_err());
+    }
+
+    #[test]
+    fn check_status_validates_post_trim() {
+        // Leading/trailing whitespace is trimmed before the length
+        // check — a borderline-legal payload with spaces around it
+        // still passes.
+        let padded = format!("  {}  ", "a".repeat(STATUS_MAX_CHARS));
+        assert!(check_status_text(&padded).is_ok());
+    }
 }