Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(security): honor detect-secrets allowlist pragmas (#3246)

Emre K.03f03f57fbb0
diff --git a/docs/layers/SECURITY.md b/docs/layers/SECURITY.md
index 7cb5803a8..3ae3615b4 100644
--- a/docs/layers/SECURITY.md
+++ b/docs/layers/SECURITY.md
@@ -193,6 +193,14 @@ lists. There is no next-line or whole-file form. A silenced finding never fails
the gate and is still reported as suppressed. Exact rules are in the
[CLI reference](../reference/CLI_REFERENCE.md#repowise-security-check-revspec).
+The scanner also honours detect-secrets comments: `# pragma: allowlist secret`
+on the finding's line, or `# pragma: allowlist nextline secret` on the line
+above it (`//` and `/*` comments work too). Unlike the Repowise gate-only
+marker, allowlisted secret findings are excluded from the stored scan and
+the gate; they do not appear in the gate's suppressed count. Non-secret
+findings on the same line still report. A marker inside a string does not
+silence a finding.
+
### Baselines
`--write-baseline FILE` records the change's findings in a committed JSON file
diff --git a/packages/core/src/repowise/core/analysis/security_gate.py b/packages/core/src/repowise/core/analysis/security_gate.py
index f981b2e84..14f28429d 100644
--- a/packages/core/src/repowise/core/analysis/security_gate.py
+++ b/packages/core/src/repowise/core/analysis/security_gate.py
@@ -41,7 +41,14 @@ from repowise.core.support_paths import DOC_EXTENSIONS
from .change_risk.features import GIT_TIMEOUT_SECONDS, _git, revspec_head, split_revspec
from .changed_lines import DIFF_ARGS, parse_unified_diff
-from .security_scan import SECRET_KINDS, masked_snippet, scan_source, source_lines
+from .security_scan import (
+ SECRET_KINDS,
+ _detect_secrets_allowlisted,
+ _mask_comments_and_strings,
+ masked_snippet,
+ scan_source,
+ source_lines,
+)
#: Severities from least to most severe; ``--fail-on`` names the lowest that fails.
SEVERITIES = ("low", "med", "high")
@@ -440,7 +447,17 @@ def _scoped(
if (entry := _builtin_entry(text, f, lines, secrets_only)) is not None
]
custom, long_lines = _custom_entries(text, lines)
- return [_row(text, entry) for entry in [*found, *custom]], long_lines
+ uncommented = (
+ source_lines(_mask_comments_and_strings(source, strings=False))
+ if "pragma: allowlist" in source
+ else text.lines
+ )
+ return [
+ _row(text, entry)
+ for entry in [*found, *custom]
+ if (entry[1] not in SECRET_KINDS and not entry[1].startswith("custom:"))
+ or not _detect_secrets_allowlisted(text.lines, uncommented, entry[0])
+ ], long_lines
#: ``(line, kind, severity, snippet shown, fingerprint)``
diff --git a/packages/core/src/repowise/core/analysis/security_scan.py b/packages/core/src/repowise/core/analysis/security_scan.py
index 6d5ddaf7f..44b4b42e3 100644
--- a/packages/core/src/repowise/core/analysis/security_scan.py
+++ b/packages/core/src/repowise/core/analysis/security_scan.py
@@ -43,6 +43,30 @@ logger = logging.getLogger(__name__)
# happens to edit.
SECURITY_SCANNER_VERSION = 2
+# detect-secrets allows a marker on the finding's line or the line above.
+_ALLOWLIST_SAME_RE = re.compile(r"(?:#|//|/\*)\s*pragma:\s*allowlist\s+secret\b")
+_ALLOWLIST_NEXT_RE = re.compile(
+ r"(?:#|//|/\*)\s*pragma:\s*allowlist\s+nextline\s+secret\b"
+)
+
+
+def _detect_secrets_allowlisted(
+ lines: list[str], uncommented: list[str], lineno: int
+) -> bool:
+ """Whether a pragma in a comment (not a string) allowlists this line."""
+
+ def marked(index: int, pattern: re.Pattern[str]) -> bool:
+ return any(
+ uncommented[index][m.start() : m.end()] != m.group()
+ for m in pattern.finditer(lines[index])
+ )
+
+ return 0 < lineno <= len(lines) and (
+ marked(lineno - 1, _ALLOWLIST_SAME_RE)
+ or (lineno > 1 and marked(lineno - 2, _ALLOWLIST_NEXT_RE))
+ )
+
+
_CREDENTIAL_EXACT_PLACEHOLDERS: frozenset[str] = frozenset({"password", "changeit"})
_CREDENTIAL_SUBSTRING_PLACEHOLDERS: tuple[str, ...] = (
@@ -1000,7 +1024,16 @@ def scan_source(file_path: str, source: str, symbols: Iterable[Any] = ()) -> lis
}
)
- return findings
+ uncommented = (
+ source_lines(_mask_comments_and_strings(source, strings=False))
+ if "pragma: allowlist" in source
+ else lines
+ )
+ return [
+ f for f in findings
+ if f["kind"] not in SECRET_KINDS
+ or not _detect_secrets_allowlisted(lines, uncommented, f["line"])
+ ]
def scan_source_map(
diff --git a/tests/unit/analysis/test_security_gate_extras.py b/tests/unit/analysis/test_security_gate_extras.py
index 28ca79270..3d15152eb 100644
--- a/tests/unit/analysis/test_security_gate_extras.py
+++ b/tests/unit/analysis/test_security_gate_extras.py
@@ -96,6 +96,24 @@ def test_a_bare_marker_silences_every_kind_on_its_line(repo):
(suppressed,) = scan.suppressed
assert (suppressed["kind"], suppressed["line_number"]) == ("aws_access_key", 1)
+def test_detect_secrets_pragmas_do_not_enter_gate_or_suppressed_report(repo):
+ _write(
+ repo,
+ "cfg.py",
+ f"AWS = '{KEY}'; eval(payload) # pragma: allowlist secret\n"
+ "# pragma: allowlist nextline secret\n"
+ f"AWS = '{KEY}'\n"
+ f"AWS = '{KEY}'\n",
+ )
+ _commit(repo, "add")
+ scan = _scan(repo, patterns=(_pattern(regex=r"AKIA[A-Z]{16}"),))
+ assert [(f["kind"], f["line_number"]) for f in scan.findings] == [
+ ("eval_call", 1),
+ ("aws_access_key", 4),
+ ("custom:internal_token", 4),
+ ]
+ assert scan.suppressed == []
+
def test_a_kind_scoped_marker_silences_only_the_kinds_it_names(repo):
_write(
diff --git a/tests/unit/analysis/test_security_scan.py b/tests/unit/analysis/test_security_scan.py
index 25def5de4..b7f8c09db 100644
--- a/tests/unit/analysis/test_security_scan.py
+++ b/tests/unit/analysis/test_security_scan.py
@@ -99,6 +99,32 @@ class TestScanFile:
findings = asyncio.run(scanner.scan_file("b.py", CLEAN.decode(), symbols=[]))
assert findings == []
+ @pytest.mark.parametrize(
+ "allowed",
+ [
+ 'password = "actualcredential123" # pragma: allowlist secret\n',
+ '# pragma: allowlist nextline secret\npassword = "actualcredential123"\n',
+ ],
+ )
+ def test_detect_secrets_pragmas_exclude_findings_before_storage(self, allowed: str) -> None:
+ source = allowed + 'password = "anothercredential456"\n'
+ findings = scan_source("config.py", source)
+ assert [(f["kind"], f["line"]) for f in findings] == [
+ ("hardcoded_password", len(allowed.splitlines()) + 1)
+ ]
+
+ def test_detect_secrets_pragma_inside_a_string_does_not_allowlist(self) -> None:
+ source = 'password = "actualcredential123 # pragma: allowlist secret"\n'
+ assert [(f["kind"], f["line"]) for f in scan_source("config.py", source)] == [
+ ("hardcoded_password", 1)
+ ]
+
+ def test_detect_secrets_pragma_does_not_hide_non_secret_calls(self) -> None:
+ source = 'password = "actualcredential123"; eval(payload) # pragma: allowlist secret\n'
+ assert [(f["kind"], f["line"]) for f in scan_source("config.py", source)] == [
+ ("eval_call", 1)
+ ]
+
@pytest.mark.parametrize(
("source", "expected"),
[