address review

Signed-off-by: David Korczynski <david@adalogics.com>
google · Jul 15, 2024 · 19fc62e · 19fc62e
1 parent aa59a04
commit 19fc62e
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 3 deletions.
diff --git a/experiment/evaluator.py b/experiment/evaluator.py
@@ -263,7 +263,7 @@ def extend_build_with_corpus(self, ai_binary, target_path,
     generated_project_path = os.path.join(oss_fuzz_checkout.OSS_FUZZ_DIR,
                                           'projects',
                                           generated_oss_fuzz_project)
-    generated_corp = corpus_generator.generate_corpus(
+    generated_corp = corpus_generator.get_corpus_generator_script(
         ai_binary, self.builder_runner.fixer_model_name, target_path,
         self.benchmark)
 

diff --git a/llm_toolkit/corpus_generator.py b/llm_toolkit/corpus_generator.py
@@ -24,14 +24,19 @@
 from llm_toolkit import prompt_builder
 
 
-def generate_corpus(
+def get_corpus_generator_script(
     ai_binary: str,
     fixer_model_name: str,
     target_harness_path: str,
     benchmark: Benchmark,
 ) -> str:
   """Uses LLMs to generate a python script that will create a seed corpus for a
-  harness."""
+  harness.
+
+  The script generated is purely generated and should be considered untrusted
+  in the general sense. OSS-Fuzz-gen already executes arbitrary code since
+  OSS-Fuzz-gen executes arbitrary open source projects with no checking on
+  what code is committed to the given projects."""
   corpus_model = models.LLM.setup(
       ai_binary=ai_binary,
       name=fixer_model_name,