diff --git a/CHANGELOG.md b/CHANGELOG.md
index 578342c1..dc7ddab1 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,11 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ## [Unreleased]
 
+### Fixed
+- Filter metrics not appearing in TUI Resources panel when using native tool_use providers (#480)
+- Output filter matchers not matching compound shell commands like `cd /path && cargo test 2>&1 | tail` (#481)
+- Duplicate `ToolEvent::Completed` emission in shell executor before filtering was applied (#480)
+
 ### Added
 - Syntax-highlighted diff view for write/edit tool output in TUI (#451)
   - Diff rendering with green/red backgrounds for added/removed lines
diff --git a/crates/zeph-tools/src/filter/mod.rs b/crates/zeph-tools/src/filter/mod.rs
index 1db617e4..d4abe0c0 100644
--- a/crates/zeph-tools/src/filter/mod.rs
+++ b/crates/zeph-tools/src/filter/mod.rs
@@ -68,6 +68,11 @@ pub enum CommandMatcher {
 impl CommandMatcher {
     #[must_use]
     pub fn matches(&self, command: &str) -> bool {
+        self.matches_single(command)
+            || extract_last_command(command).is_some_and(|last| self.matches_single(last))
+    }
+
+    fn matches_single(&self, command: &str) -> bool {
         match self {
             Self::Exact(s) => command == *s,
             Self::Prefix(s) => command.starts_with(s),
@@ -77,6 +82,29 @@ impl CommandMatcher {
     }
 }
 
+/// Extract the last command segment from compound shell expressions
+/// like `cd /path && cargo test` or `cmd1 ; cmd2`. Strips trailing
+/// redirections and pipes (e.g. `2>&1 | tail -50`).
+fn extract_last_command(command: &str) -> Option<&str> {
+    let last = command
+        .rsplit("&&")
+        .next()
+        .or_else(|| command.rsplit(';').next())?;
+    let last = last.trim();
+    if last == command.trim() {
+        return None;
+    }
+    // Strip trailing pipe chain and redirections: take content before first `|` or `2>`
+    let last = last.split('|').next().unwrap_or(last);
+    let last = last.split("2>").next().unwrap_or(last);
+    let trimmed = last.trim();
+    if trimmed.is_empty() {
+        None
+    } else {
+        Some(trimmed)
+    }
+}
+
 impl std::fmt::Debug for CommandMatcher {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         match self {
@@ -754,6 +782,38 @@ extra_patterns = ["TODO: security review"]
         assert!(!m.matches("goodbye"));
     }
 
+    #[test]
+    fn command_matcher_compound_cd_and() {
+        let m = CommandMatcher::Prefix("cargo ");
+        assert!(m.matches("cd /some/path && cargo test --workspace --lib"));
+        assert!(m.matches("cd /path && cargo clippy --workspace -- -D warnings 2>&1"));
+    }
+
+    #[test]
+    fn command_matcher_compound_with_pipe() {
+        let m = CommandMatcher::Custom(Box::new(|cmd| cmd.split_whitespace().any(|t| t == "test")));
+        assert!(m.matches("cd /path && cargo test --workspace --lib 2>&1 | tail -80"));
+    }
+
+    #[test]
+    fn command_matcher_compound_no_false_positive() {
+        let m = CommandMatcher::Exact("ls");
+        assert!(!m.matches("cd /path && cargo test"));
+    }
+
+    #[test]
+    fn extract_last_command_basic() {
+        assert_eq!(
+            extract_last_command("cd /path && cargo test --lib"),
+            Some("cargo test --lib")
+        );
+        assert_eq!(
+            extract_last_command("cd /p && cargo clippy 2>&1 | tail -20"),
+            Some("cargo clippy")
+        );
+        assert!(extract_last_command("cargo test").is_none());
+    }
+
     // FilterConfidence derives
     #[test]
     fn filter_confidence_derives() {
diff --git a/docs/src/guide/tools.md b/docs/src/guide/tools.md
index ce0f9d02..c3645a93 100644
--- a/docs/src/guide/tools.md
+++ b/docs/src/guide/tools.md
@@ -115,6 +115,10 @@ Stale overflow files older than 24 hours are cleaned up automatically on startup
 
 Before tool output reaches the LLM context, it passes through a command-aware filter pipeline that strips noise and reduces token consumption. Filters are matched by command pattern and composed in sequence.
 
+### Compound Command Matching
+
+LLMs often generate compound shell expressions like `cd /path && cargo test 2>&1 | tail -80`. Filter matchers automatically extract the last command segment after `&&` or `;` separators and strip trailing pipes and redirections before matching. This means `cd /Users/me/project && cargo clippy --workspace -- -D warnings 2>&1` correctly matches the `ClippyFilter` — no special configuration needed.
+
 ### Built-in Filters
 
 | Filter | Matches | What it removes |
@@ -141,7 +145,7 @@ Each filter reports a confidence level:
 | `Partial` | Heuristic match; some content may have been over-filtered |
 | `Fallback` | Pattern matched but output structure was unexpected |
 
-When multiple filters compose in a pipeline, the worst confidence across stages is propagated. Confidence distribution is tracked in [TUI filter metrics](tui.md#filter-metrics).
+When multiple filters compose in a pipeline, the worst confidence across stages is propagated. Confidence distribution is tracked in the [TUI Resources panel](tui.md#confidence-levels-explained) as `F/P/B` counters.
 
 ### Inline Filter Stats (CLI)
 
diff --git a/docs/src/guide/tui.md b/docs/src/guide/tui.md
index 7a5635cf..cff29381 100644
--- a/docs/src/guide/tui.md
+++ b/docs/src/guide/tui.md
@@ -204,10 +204,24 @@ Confidence: F/6 P/2 B/0
 | `hit rate` | Percentage of commands where output was actually reduced |
 | `saved tokens` | Cumulative estimated tokens saved (`chars_saved / 4`) |
 | `%` | Token savings as a fraction of raw token volume |
-| `F/P/B` | Confidence distribution: Full / Partial / Fallback counts |
+| `F/P/B` | Confidence distribution: Full / Partial / Fallback counts (see below) |
 
 The filter section only appears when `filter_applications > 0` — it is hidden when no commands have been filtered.
 
+#### Confidence Levels Explained
+
+Each filter reports how confident it is in the result. The `Confidence: F/1 P/0 B/3` line shows cumulative counts across all filtered commands:
+
+| Level | Abbreviation | When assigned | What it means for the output |
+|-------|-------------|---------------|------------------------------|
+| **Full** | `F` | Filter recognized the output structure completely (e.g. `cargo test` with standard `test result:` summary) | Output is reliably compressed — no useful information lost |
+| **Partial** | `P` | Filter matched the command but output had unexpected sections mixed in (e.g. warnings interleaved with test results) | Most noise removed, but some relevant content may have been stripped — inspect if results look incomplete |
+| **Fallback** | `B` | Command pattern matched but output structure was unrecognized (e.g. `cargo audit` matched a cargo-prefix filter but has no dedicated handler) | Output returned unchanged or with minimal sanitization only (ANSI stripping, blank line collapse) |
+
+**Example:** `Confidence: F/1 P/0 B/3` means 1 command was filtered with Full confidence (e.g. `cargo test` — 99% savings) and 3 commands fell through to Fallback (e.g. `cargo audit`, `cargo doc`, `cargo tree` — matched the filter pattern but output was passed through as-is).
+
+When multiple filters compose in a [pipeline](tools.md#output-filter-pipeline), the worst confidence across stages is propagated. A `Full` + `Partial` composition yields `Partial`.
+
 ## Deferred Model Warmup
 
 When running with Ollama (or an orchestrator with Ollama sub-providers), model warmup is deferred until after the TUI interface renders. This means: