feat(controls): enrich semantic retrieval query with finding intent + C5 live test

Tuning from the C5 live run against the real 2,882-control master corpus. The
semantic pass retrieved on the code region alone, so two findings in one file
(overlapping windows, both md5/password tokens) collapsed onto the SAME controls —
a brute-force finding wrongly matched password-hashing controls.

Fix: build the retrieval query from the finding's title + description + region, so
retrieval keys on what the finding is *about*. The raw region still goes to the
judge for snippet grounding.

Verified live (same fixture, cached corpus index):
- 'Weak password hash (md5)'      -> mc-23149 (eliminate weak unsalted hashes) now RANK 1
- 'Login without brute-force prot' -> newly surfaces mc-19984 (brute_force_protection)
                                      + mc-23186 (account_lockout) — the correct controls,
                                      absent under region-only retrieval.

Also commits the gated live regression test (tests/c5_semantic_live.rs, #[ignore]d,
not run by CI's --lib): ingest-only + full semantic-stamping checks against api-dev.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

compliance-agent/src/controls/scan_triage.rs

@@ -234,13 +234,22 @@ pub async fn semantic_stamp_findings(
         let Some(region) = fetch_region(repo_path, &file, line) else {
             continue;
         };
-        let region_emb = match llm.embed(vec![region.content.clone()]).await {
+        // Retrieve on the finding's intent + the code, not the region alone: two
+        // findings in one file share overlapping windows and otherwise embed alike,
+        // collapsing onto the same controls. The finding's title/description carry
+        // the discriminating signal (e.g. "brute-force protection" vs "weak hash").
+        // The raw `region` still goes to the judge for snippet grounding.
+        let query = format!(
+            "{}\n{}\n\n{}",
+            finding.title, finding.description, region.content
+        );
+        let query_emb = match llm.embed(vec![query]).await {
             Ok(mut embs) => match embs.pop() {
                 Some(v) => v,
                 None => continue,
             },
             Err(e) => {
-                tracing::warn!(error = %e, "region embed failed; skipping finding");
+                tracing::warn!(error = %e, "query embed failed; skipping finding");
                 continue;
             }
         };
@@ -248,7 +257,7 @@ pub async fn semantic_stamp_findings(
             .check(
                 &index,
                 &region,
-                &region_emb,
+                &query_emb,
                 SEMANTIC_TOP_K,
                 &finding.repo_id,
             )

compliance-agent/src/controls/semantic.rs

@@ -22,18 +22,20 @@ impl<J: ControlJudge> SemanticControlChecker<J> {
         Self { judge }
     }
 
-    /// Map a code region to the controls it violates. `region_embedding` is the
-    /// region's embedding (the caller computes it via the LLM); the top-`k`
-    /// nearest controls in `index` are judged and grounded.
+    /// Map a code region to the controls it violates. `query_embedding` is the
+    /// caller-supplied retrieval embedding — typically the finding's intent
+    /// (title/description) plus the region, so retrieval keys on what the finding
+    /// is *about*, not just the ambient code. The top-`k` nearest controls in
+    /// `index` are then judged against the raw `region` and grounded.
     pub async fn check(
         &self,
         index: &ControlIndex,
         region: &CandidateRegion,
-        region_embedding: &[f64],
+        query_embedding: &[f64],
         k: usize,
         repo_id: &str,
     ) -> Vec<Finding> {
-        let candidates = index.nearest(region_embedding, k);
+        let candidates = index.nearest(query_embedding, k);
         let mut findings = Vec::new();
         for spec in &candidates {
             let verdict = self.judge.judge(spec, region).await;

compliance-agent/src/controls/surface.rs

@@ -47,6 +47,17 @@ pub const SURFACES: &[Surface] = &[
             "ratelimit",
         ],
     },
+    Surface {
+        control_id: "cra-ai-12", // Rollenbasierte Autorisierung (RBAC)
+        terms: &[
+            "authorize",
+            "permission",
+            "role",
+            "rbac",
+            "require_role",
+            "has_role",
+        ],
+    },
     Surface {
         control_id: "cra-ai-24", // Security-Logging
         terms: &["login", "authorize", "permission", "role", "admin", "audit"],
@@ -196,10 +207,11 @@ mod tests {
 
     #[test]
     fn surfaces_cover_the_absence_based_controls() {
-        assert_eq!(SURFACES.len(), 7);
+        assert_eq!(SURFACES.len(), 8);
         for id in [
             "cra-ai-6",
             "cra-ai-11",
+            "cra-ai-12",
             "cra-ai-24",
             "cra-ai-27",
             "cra-ai-28",

compliance-agent/src/llm/embedding.rs

@@ -22,6 +22,11 @@ struct EmbeddingData {
     index: usize,
 }
 
+/// Max inputs per embedding request. The bge/OpenAI-like backends cap the input
+/// array (bge-multilingual-gemma2 rejects >25 with "batch size overflow"), so we
+/// chunk larger corpora — a whole control catalog (~1.8k) would otherwise 500.
+const EMBED_BATCH_SIZE: usize = 16;
+
 // ── Embedding implementation ───────────────────────────────────
 
 impl LlmClient {
@@ -29,8 +34,21 @@ impl LlmClient {
         &self.embed_model
     }
 
-    /// Generate embeddings for a batch of texts
+    /// Generate embeddings for a batch of texts, chunking into backend-sized
+    /// requests and preserving input order across chunks.
     pub async fn embed(&self, texts: Vec<String>) -> Result<Vec<Vec<f64>>, AgentError> {
+        if texts.is_empty() {
+            return Ok(Vec::new());
+        }
+        let mut out = Vec::with_capacity(texts.len());
+        for chunk in texts.chunks(EMBED_BATCH_SIZE) {
+            out.extend(self.embed_batch(chunk.to_vec()).await?);
+        }
+        Ok(out)
+    }
+
+    /// Embed one backend-sized batch (≤ [`EMBED_BATCH_SIZE`]) in a single request.
+    async fn embed_batch(&self, texts: Vec<String>) -> Result<Vec<Vec<f64>>, AgentError> {
         let url = format!("{}/v1/embeddings", self.base_url.trim_end_matches('/'));
 
         let request_body = EmbeddingRequest {
@@ -72,3 +90,33 @@ impl LlmClient {
         Ok(data.into_iter().map(|d| d.embedding).collect())
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use secrecy::SecretString;
+
+    fn client() -> LlmClient {
+        LlmClient::new(
+            "http://unused".into(),
+            SecretString::from(String::new()),
+            "m".into(),
+            "e".into(),
+        )
+    }
+
+    #[tokio::test]
+    async fn empty_input_makes_no_request() {
+        // Must short-circuit before any HTTP call (base_url is unroutable).
+        let out = client().embed(Vec::new()).await.unwrap();
+        assert!(out.is_empty());
+    }
+
+    #[test]
+    fn batch_size_is_within_backend_cap() {
+        assert!(
+            EMBED_BATCH_SIZE <= 25,
+            "must stay under the bge 25-input cap"
+        );
+    }
+}

compliance-agent/tests/c5_semantic_live.rs

@@ -0,0 +1,145 @@
+//! C5 live verification — the semantic master-controls path end to end against the
+//! deployed api-dev catalog. Ignored (hits api-dev + LiteLLM). Run explicitly:
+//!
+//!   set -a; . ./.env; set +a
+//!   BREAKPILOT_BASE_URL=https://api-dev.breakpilot.ai \
+//!     cargo test -p compliance-agent --test c5_semantic_live -- --ignored --nocapture
+//!
+//! Pulls the live master-controls catalog, embeds the corpus (chunked), then for a
+//! couple of real vulnerable findings retrieves the nearest master controls and
+//! grounded-judges them, stamping master-control refs.
+
+mod common;
+
+use std::sync::Arc;
+
+use compliance_agent::llm::LlmClient;
+use compliance_core::config::BreakpilotConfig;
+use compliance_core::models::finding::{Finding, Severity};
+use compliance_core::models::scan::ScanType;
+use secrecy::SecretString;
+
+fn env(k: &str) -> String {
+    std::env::var(k).unwrap_or_else(|_| panic!("env {k} must be set for the live C5 test"))
+}
+
+fn mk_finding(file: &str, line: u32, title: &str) -> Finding {
+    let mut f = Finding::new(
+        "repo-c5".into(),
+        format!("{file}:{line}"),
+        "semgrep".into(),
+        ScanType::Sast,
+        title.into(),
+        title.into(),
+        Severity::High,
+    );
+    f.file_path = Some(file.into());
+    f.line_number = Some(line);
+    f
+}
+
+#[tokio::test]
+#[ignore = "live: requires deployed api-dev master-controls (fetch+parse only, no LLM)"]
+async fn c5_ingest_master_controls_catalog() {
+    use compliance_agent::controls::OscalControlsProvider;
+
+    let provider = OscalControlsProvider::new(
+        reqwest::Client::new(),
+        env("BREAKPILOT_BASE_URL"),
+        None,
+        std::env::temp_dir().join("c5-ingest-snap"),
+    );
+    let doc = provider
+        .load_master_controls()
+        .await
+        .expect("pull + parse master-controls catalog");
+    let controls = doc.to_controls();
+    println!(
+        "\n=== C5 ingest: {} master controls parsed ===",
+        controls.len()
+    );
+    for c in controls.iter().take(4) {
+        let text: String = c.text.chars().take(90).collect();
+        println!("  {} | {} | {}", c.id, c.title, text);
+    }
+    assert!(
+        !controls.is_empty(),
+        "expected a non-empty master-control corpus"
+    );
+}
+
+#[tokio::test]
+#[ignore = "live: requires deployed api-dev master-controls + LiteLLM"]
+async fn c5_semantic_stamps_master_control_refs() {
+    let llm = Arc::new(LlmClient::new(
+        env("LITELLM_URL"),
+        SecretString::from(env("LITELLM_API_KEY")),
+        env("LITELLM_MODEL"),
+        env("LITELLM_EMBED_MODEL"),
+    ));
+
+    let mut config = common::dev_config("mongodb://unused".into(), "c5".into());
+    let snapshot = std::env::temp_dir().join("c5-oscal-snap");
+    config.breakpilot = BreakpilotConfig {
+        base_url: Some(env("BREAKPILOT_BASE_URL")),
+        token: None,
+        snapshot_dir: snapshot.to_string_lossy().into_owned(),
+        semantic_mapping: true,
+        grounded_control_checks: false,
+    };
+
+    // Fixture repo with recognizable code-checkable surfaces.
+    let repo = std::env::temp_dir().join("c5-fixture-repo");
+    let _ = std::fs::remove_dir_all(&repo);
+    std::fs::create_dir_all(repo.join("app")).expect("mkdir");
+    std::fs::write(
+        repo.join("app/auth.py"),
+        concat!(
+            "import hashlib\n",
+            "\n",
+            "def store_password(user, password):\n",
+            "    # weak, unsalted password hashing\n",
+            "    digest = hashlib.md5(password.encode()).hexdigest()\n",
+            "    db.save(user, digest)\n",
+            "\n",
+            "@app.route('/login', methods=['POST'])\n",
+            "def login():\n",
+            "    u = request.form['username']\n",
+            "    p = request.form['password']\n",
+            "    return 'ok' if check(u, p) else ('bad', 401)\n",
+        ),
+    )
+    .expect("write fixture");
+
+    let mut findings = vec![
+        mk_finding("app/auth.py", 5, "Weak password hash (md5, unsalted)"),
+        mk_finding(
+            "app/auth.py",
+            9,
+            "Login endpoint without brute-force protection",
+        ),
+    ];
+
+    let tagged =
+        compliance_agent::controls::semantic_stamp_findings(&config, llm, &repo, &mut findings)
+            .await;
+
+    println!("\n=== C5 semantic master-controls stamping ===");
+    for f in &findings {
+        println!(
+            "  {:50} {}:{:?} -> {:?}",
+            f.title,
+            f.file_path.as_deref().unwrap_or(""),
+            f.line_number,
+            f.control_refs
+        );
+    }
+    println!("findings that gained >=1 master-control ref: {tagged}");
+    let _ = std::fs::remove_dir_all(&repo);
+
+    // Live corpus — assert only that the path runs and stamps at least one ref.
+    assert!(
+        tagged >= 1,
+        "expected at least one finding to gain a master-control ref"
+    );
+}

control-map/data/cra_control_map.json

@@ -25,29 +25,29 @@
       "control": "cra-ai-2",
       "title": "Minimale Angriffsflaeche",
       "scans": [],
-      "note": "code-checkable but no off-the-shelf tool digs it out — author a detector (custom semgrep rule / check)",
-      "status": "needs_tooling"
+      "note": "design property (minimal attack surface) — not derivable from local code patterns; architecture/threat-model review",
+      "status": "not_code_checkable"
     },
     {
       "control": "cra-ai-3",
       "title": "Sichere Systemarchitektur",
       "scans": [],
-      "note": "code-checkable but no off-the-shelf tool digs it out — author a detector (custom semgrep rule / check)",
-      "status": "needs_tooling"
+      "note": "design property (secure system architecture) — architecture review, not statically code-checkable",
+      "status": "not_code_checkable"
     },
     {
       "control": "cra-ai-4",
       "title": "Least-Privilege-Prinzip",
       "scans": [],
-      "note": "code-checkable but no off-the-shelf tool digs it out — author a detector (custom semgrep rule / check)",
-      "status": "needs_tooling"
+      "note": "design property (least-privilege) — deployment/IAM & architecture review, not a local code pattern",
+      "status": "not_code_checkable"
     },
     {
       "control": "cra-ai-5",
       "title": "Manipulationsschutz",
       "scans": [],
-      "note": "code-checkable but no off-the-shelf tool digs it out — author a detector (custom semgrep rule / check)",
-      "status": "needs_tooling"
+      "note": "design property (tamper protection) — hardware/runtime & operational control, not statically code-checkable",
+      "status": "not_code_checkable"
     },
     {
       "control": "cra-ai-6",
@@ -146,7 +146,7 @@
       "control": "cra-ai-12",
       "title": "Rollenbasierte Autorisierung",
       "scans": [],
-      "note": "code-checkable but no off-the-shelf tool digs it out — author a detector (custom semgrep rule / check)",
+      "note": "absence-based — no syntactic pattern; covered by the grounded surface check (retrieve surface + LLM judge), gated (BREAKPILOT_GROUNDED_CHECKS) pending live tuning",
       "status": "needs_tooling"
     },
     {

control-map/src/lib.rs

@@ -214,6 +214,27 @@ mod tests {
         assert!(hits.iter().any(|c| c.control == "cra-ai-1"));
     }
 
+    #[test]
+    fn coverage_reflects_the_b_track_split() {
+        let s = ControlMap::cra().unwrap().summary();
+        // 9 already tool-covered + B1's 4 custom-semgrep controls.
+        assert_eq!(s.covered, 13);
+        // The 8 grounded surface controls stay needs_tooling until live-tuned.
+        assert_eq!(s.needs_tooling, 8);
+        // B3 marked the 4 pure-architectural controls not code-checkable.
+        assert_eq!(s.not_code_checkable, 19);
+    }
+
+    #[test]
+    fn architectural_controls_are_not_code_checkable() {
+        let map = ControlMap::cra().unwrap();
+        for id in ["cra-ai-2", "cra-ai-3", "cra-ai-4", "cra-ai-5"] {
+            let c = map.coverage(id).unwrap();
+            assert_eq!(c.status, Coverage::NotCodeCheckable, "{id}");
+            assert!(c.scans.is_empty(), "{id} should carry no scan bindings");
+        }
+    }
+
     #[test]
     fn custom_rule_controls_do_not_bind_by_broad_cwe() {
         let map = ControlMap::cra().unwrap();