apunkt/nyx
1
0
Fork 0
mirror of https://github.com/elicpeter/nyx.git synced 2026-06-21 20:18:06 +02:00
Code Issues Projects Releases Packages Wiki Activity

feat(dynamic): improve sandbox hardening and build caching

Browse source
This commit is contained in:
elipeter 2026-06-03 12:26:10 -05:00
parent 7468d2214b
commit 20093972a9
8 changed files with 345 additions and 45 deletions
Download patch file Download diff file Expand all files Collapse all files

11
tests/eval_corpus/tabulate.py
View file

@ -454,6 +454,17 @@ def main() -> int:
scan_data = load_json(args.scan)
findings = scan_data if isinstance(scan_data, list) else scan_data.get("findings", [])
# Score only Security-category findings against the security ground truth.
# Reliability defects (resource leaks, error-handling fallthrough) and
# Quality findings are real bugs but not the injection / crypto / auth
# vulns the corpus ground truth enumerates, so counting them as security
# false-positives is a category error that wrecks precision with pure
# noise. Findings with no explicit category (legacy fixtures) default to
# Security and are kept.
findings = [
f for f in findings
if f.get("category", "Security") not in ("Reliability", "Quality")
]
if lang_filter:
findings = [f for f in findings if lang_of(f) in lang_filter]