7 Commits
Author SHA1 Message Date
Martin Vogel 72c5bdba89 fix(security-strings): allowlist 'telnet' from rst grammar URI schemes
The rst tree-sitter grammar (added in the 89-grammar bump) contains a
valid_schemas[] array listing URI schemes (http, https, ftp, mailto,
telnet, ssh) in vendored/grammars/rst/tree_sitter_rst/chars.c. The
'telnet' string ends up in the binary's string table and tripped the
dangerous-command detector, blocking smoke on every platform.

Add an allowlist mechanism for known-benign matches with a comment
pointing at the source file, so future false positives can be
documented the same way.
2026-05-04 22:04:39 +02:00
Martin VogelanddLo999 6f268d0f91 Respect nested .gitignore files during indexing, security audit all variants
Nested .gitignore support (fixes #178):
- Load per-subdirectory .gitignore during walk, match paths relative to
  the gitignore's directory via local_rel_path()
- Root and nested gitignores stack independently
- Owned gitignores collected and freed at walk end (avoids use-after-free
  from borrowed pointers on the iterative stack)

Security:
- Run security-strings/install/network + ClamAV + Windows Defender on
  ALL binary variants (standard + UI), not just standard
- Whitelist UI bundle URLs (React, Three.js, Google Fonts, Tailwind, W3C)

Co-Authored-By: dLo999 <dLo999@users.noreply.github.com>
2026-04-06 19:44:16 +02:00
Martin Vogel 9e1e30ded0 Allow toolchain URLs in binary string audit (static build artifacts)
gcc/glibc embed bug tracker URLs (bugs.launchpad.net, gcc.gnu.org,
sourceware.org) into statically-linked binaries. These are compiler
artifacts, not our code.
2026-04-06 17:47:04 +02:00
Martin Vogel 8967d7cd41 Remove AV-triggering words from token vocabulary, revert audit allowlist
Strip 11 tokens (wget, curl, netcat, ncat, telnet, passwd, shadow,
exploit, hack, inject, malware) from Nomic vocabulary. These fall back
to sparse random vectors — negligible quality impact. Removes all
security audit exceptions: zero allowlists, zero suppressions.
2026-04-06 16:12:02 +02:00
Martin Vogel 3f7f7eeab7 Fix binary string audit: exclude bare token vocabulary matches
The embedded Nomic code token vocabulary (40K tokens) includes words like
"wget" as code tokens. Filter out bare single-word matches (2-10 lowercase
chars) since real dangerous strings appear in command context, not as
standalone vocabulary entries.
2026-04-06 15:40:11 +02:00
Martin Vogel 93d332d1e0 Fix security audit false positives in CI
- UI audit (L6): allowlist bundled framework URLs in dist/ (React error
  URLs, W3C namespace URIs, Three.js credits, Google Fonts, Tailwind).
  These are embedded by npm deps during Vite build, not our code.
- Binary strings (L2): skip URLs shorter than 15 chars — Windows binary
  has byte sequences that strings(1) interprets as "https://H9" etc.
- Allow Google Fonts <link> in HTML (loaded by index.html for Inter/
  JetBrains Mono fonts).
2026-03-20 18:47:55 +01:00
Martin Vogel cd1417427c Add 8-layer security test suite + hardening
Code-level defenses:
- cbm_validate_shell_arg(): reject shell metacharacters before popen/system
- SQLite authorizer: block ATTACH/DETACH at engine level
- CORS localhost-only origin reflection (replaces wildcard *)
- Path containment: realpath() check in get_code_snippet
- process-kill restricted to server-spawned PIDs
- SHA256 checksum verification in update command

Security audit scripts (8 layers):
- L1: Static allow-list for dangerous calls + URLs
- L2: Binary string audit (URLs, payloads, credentials)
- L3: Network egress monitoring via strace (Linux)
- L4: Install output path + content validation
- L5: Smoke test hardening (clean shutdown, residual procs)
- L6: Graph UI audit (external domains, CORS, binding)
- L7: MCP robustness (23 adversarial JSON-RPC payloads)
- L8: Vendored integrity (checksums + dangerous call scan)

CI: parallel security-static job (no build needed), binary
layers in smoke jobs per-platform. Cleanup of test fixture
dirs in clean.sh + .gitignore.
2026-03-20 18:19:15 +01:00