Add CBM_EXTRA_SKIP_DIRS env var for site-specific always-skip directories

docs/cbmignore.md

@@ -69,7 +69,12 @@ a path wins. For directories:
 1. **Built-in skip list** — `.git`, `node_modules`, `dist`, `target`,
    `vendor`, tool caches, etc. (60+ names; the fast/moderate index modes add
    more, e.g. `docs`, `examples`, `testdata`). Not overridable from any
-   ignore file today.
+   ignore file today. The hardcoded list is intentionally fixed across all
+   repos, but you can extend it per invocation with the `CBM_EXTRA_SKIP_DIRS`
+   environment variable — a comma-separated list of extra directory names to
+   skip in every mode (e.g. `CBM_EXTRA_SKIP_DIRS=".dbt,.terraform-state"`).
+   Useful for site- or repo-specific artifact directories that don't belong
+   in the universal list. Whitespace around each name is trimmed.
 2. **Repo `.gitignore`** — `<repo>/.gitignore` merged with
    `<git-common-dir>/info/exclude` (worktree-aware); later patterns win on
    conflict. Honored even when the indexed directory is not a git repo root.

src/discover/discover.c

@@ -334,6 +334,44 @@ static bool resolve_global_excludes_path(char *out, size_t out_sz) {
     return false;
 }
 
+/* ── Env-var-configurable extra skip dirs ──────────────────────────
+ * CBM_EXTRA_SKIP_DIRS: comma-separated directory names to skip, on top of
+ * the hardcoded ALWAYS_SKIP_DIRS above. Lets a caller add repo-specific or
+ * site-specific artifact dirs (e.g. ".dbt", a Terraform state dir, a prior
+ * AI-tooling output dir) per invocation, without a rebuild. Applies in
+ * every mode, same as ALWAYS_SKIP_DIRS. Not cached across calls (the list
+ * is short and this keeps the function safe to call from any thread). */
+static bool str_in_env_csv(const char *s, const char *csv) {
+    if (!s || !csv || !csv[0]) {
+        return false;
+    }
+    size_t slen = strlen(s);
+    const char *p = csv;
+    while (*p) {
+        const char *comma = strchr(p, ',');
+        size_t seg_len = comma ? (size_t)(comma - p) : strlen(p);
+
+        const char *seg = p;
+        size_t len = seg_len;
+        while (len > 0 && isspace((unsigned char)seg[0])) {
+            seg++;
+            len--;
+        }
+        while (len > 0 && isspace((unsigned char)seg[len - 1])) {
+            len--;
+        }
+
+        if (len == slen && strncmp(seg, s, len) == 0) {
+            return true;
+        }
+        if (!comma) {
+            break;
+        }
+        p = comma + 1;
+    }
+    return false;
+}
+
 /* ── Public filter functions ─────────────────────── */
 
 bool cbm_should_skip_dir(const char *dirname, cbm_index_mode_t mode) {
@@ -345,6 +383,16 @@ bool cbm_should_skip_dir(const char *dirname, cbm_index_mode_t mode) {
         return true;
     }
 
+    /* Read through cbm_safe_getenv so the value is copied into a caller-owned
+     * buffer before use — raw getenv() is mt-unsafe (pointer can be invalidated
+     * by a concurrent setenv/putenv) and mishandles UTF-8 on Windows. */
+    char extra_skip[CBM_SZ_1K];
+    if (str_in_env_csv(dirname,
+                       cbm_safe_getenv("CBM_EXTRA_SKIP_DIRS", extra_skip,
+                                       sizeof(extra_skip), NULL))) {
+        return true;
+    }
+
     /* Fast discovery applies to both MODERATE and FAST — only FULL keeps everything. */
     if (mode != CBM_MODE_FULL) {
         if (str_in_list(dirname, FAST_SKIP_DIRS)) {

tests/repro/repro_invariant_discovery_fqn.c

@@ -261,6 +261,51 @@ TEST(invariant_discovery_always_skip_dirs) {
     PASS();
 }
 
+/* ── PART A TEST — CBM_EXTRA_SKIP_DIRS env var ─────────────────────────────
+ *
+ * Verifies the env-var-configurable extra skip list (discover.c
+ * cbm_should_skip_dir -> str_in_env_csv): a name NOT in the hardcoded
+ * ALWAYS_SKIP_DIRS is skipped once listed in CBM_EXTRA_SKIP_DIRS, and goes
+ * back to being indexed once the env var is cleared — proving the behavior
+ * is genuinely env-gated, not accidentally hardcoded.
+ */
+TEST(invariant_discovery_env_extra_skip_dirs) {
+    const char *custom_dir = "dbt_target_test_fixture";
+
+    /* Baseline: not in any hardcoded list, so it must be discovered today. */
+    cbm_unsetenv("CBM_EXTRA_SKIP_DIRS");
+    int baseline = check_dir_skipped(custom_dir, CBM_MODE_FULL);
+    if (baseline < 0) {
+        printf("    SETUP-ERROR  %-32s baseline\n", custom_dir);
+        ASSERT_EQ(1, 0);
+    }
+    if (baseline == 0) {
+        printf("    UNEXPECTED   %-32s already skipped with no env var set\n", custom_dir);
+    }
+    ASSERT_GT(baseline, 0);
+
+    /* With the env var listing it (alongside noise entries, whitespace-padded
+     * to also exercise the trimming path), it must now be skipped. */
+    cbm_setenv("CBM_EXTRA_SKIP_DIRS", ".codegraph, graphify-out ,dbt_target_test_fixture,.wiki", 1);
+    int with_env = check_dir_skipped(custom_dir, CBM_MODE_FULL);
+    cbm_unsetenv("CBM_EXTRA_SKIP_DIRS");
+
+    if (with_env < 0) {
+        printf("    SETUP-ERROR  %-32s with env var\n", custom_dir);
+        ASSERT_EQ(1, 0);
+    }
+    if (with_env != 0) {
+        printf("    REGRESSION   %-32s not skipped with CBM_EXTRA_SKIP_DIRS set\n", custom_dir);
+    }
+    ASSERT_EQ(with_env, 0);
+
+    /* And clearing it restores the baseline (not permanently sticky). */
+    int after_clear = check_dir_skipped(custom_dir, CBM_MODE_FULL);
+    ASSERT_GT(after_clear, 0);
+
+    PASS();
+}
+
 /* ── PART A TEST — FAST_SKIP_DIRS table (mode != CBM_MODE_FULL) ────────────
  *
  * FAST_SKIP_DIRS entries are only skipped when mode != CBM_MODE_FULL.
@@ -794,6 +839,7 @@ SUITE(repro_invariant_discovery_fqn) {
     /* Part A — Discovery hygiene */
     RUN_TEST(invariant_discovery_control_always_found);
     RUN_TEST(invariant_discovery_always_skip_dirs);
+    RUN_TEST(invariant_discovery_env_extra_skip_dirs);
     RUN_TEST(invariant_discovery_fast_skip_dirs);
 
     /* Part B — FQN same-stem distinctness */