feat(pair-corpus-honesty): add vibe-py, vfc-js and agent-vfc slices with mechanism vocabulary

Tasks 1.1, 5.1, 5.2, 5.4 and the data half of 5.3. - mechanisms.toml: closed vocabulary v1 (20 ids) every expected row cites. - vibe-py: 35 Real-Vuln-Benchmark pairs (Apache-2.0), human provenance only because the LLM-generated repositories carry no license (escalated). - vfc-js: 17 Node fix-commit pairs harvested upstream (MIT/Apache-2.0), SecBench.js itself is not vendored (no license); sinks derived on 13/17. - agent-vfc: 29 agent-trailer fixes as catalog-candidates.toml plus REVIEW.md; the loaded catalog stays empty until a human reviews them. - datasets.toml and README document every slice and its review status. Co-Authored-By: Claude Fable 5.1 <[email protected]> Claude-Session: https://claude.ai/code/session_01T7XD7x5KonMdmGRkp63EQG
2026-10-19 04:59:28 +00:00 · 2026-09-05 10:45:42 +02:00
parent c3e841003c
commit 8ab4207618
182 changed files with 31299 additions and 0 deletions
@@ -0,0 +1,61 @@
+# Provenance: D-sorganization/Tools  (fixed).
+# repo: D-sorganization/Tools
+# commit: a96bc0502a3dcdac303e4f0152508886e5c35c0a
+# parent: e735d28b82dd1fbc42c0ae01157b38d1b4c7c864
+# commit_url: https://github.com/D-sorganization/Tools/commit/a96bc0502a3dcdac303e4f0152508886e5c35c0a
+# cve: 
+# license: MIT
+# function: _is_command_allowed
+# relpath: src/shared/python/ai/tools/cli_tools.py
+# provenance: agent
+# mechanism: source_reaches_sink
+
+    def _is_command_allowed(self, command: str) -> bool:
+        """Check if a command is allowed.
+
+        Args:
+            command: Command to check.
+
+        Returns:
+            True if allowed, False otherwise.
+        """
+        # Dangerous commands are never allowed
+        dangerous = ["rm", "sudo", "chmod", "chown", "curl", "wget", "ssh"]
+
+        # Prevent shell injection by blocking command separators/operators
+        shell_operators = ["&&", "||", ";", "|", ">", "<", "$", "`", "\n", "&"]
+        if any(op in command for op in shell_operators):
+            return False
+
+        try:
+            tokens = shlex.split(command)
+            if not tokens:
+                return False
+
+            # Verify the first token is in the allowlist
+            base_cmd = tokens[0]
+            if base_cmd not in self._allowed_commands:
+                return False
+
+            # Verify no token is a dangerous command
+            for token in tokens:
+                if token in dangerous:
+                    return False
+
+                try:
+                    # Check for absolute/relative paths (e.g., /bin/rm, ./rm)
+                    if Path(token).name in dangerous:
+                        return False
+
+                    # Check for assignments passing executables (e.g., --exec=/bin/rm)
+                    if "=" in token:
+                        val = token.split("=", 1)[1]
+                        if val in dangerous or Path(val).name in dangerous:
+                            return False
+                except Exception:
+                    pass
+
+            return True
+        except ValueError:
+            # e.g., missing closing quote
+            return False