Large language models (LLMs) have shown strong ability to synthesize formal specifications from source code. Existing approaches mainly target functional contracts and loop invariants, while often overlooking frame conditions—memory locations that may or may not be modified during execution, which are essential for complete specifications.\ \ We present an LLM‑based technique that simultaneously generates ACSL frame specifications and functional contracts for C code. The method first builds a code property graph (CPG) to extract frame‑relevant information such as variable scopes, pointer aliasing, and potential write paths. This structured data is fed to the LLM as a prompt, steering it to produce frame conditions that accurately reflect actual modifications.\ \ The resulting specifications capture region‑separation constraints and data dependencies, explicitly stating which memory regions remain unchanged and which may be updated after a function call.\ \ Experiments evaluate the approach on two benchmarks. On a set of 84 real‑world ACSL‑compliant functions, the method synthesizes error‑free frame specifications for 92.9% of the functions, a 179% relative improvement over state‑of‑the‑art techniques. On a second benchmark of 51 functions, the success rate rises from 80.4% to 96.1%. When integrated as an auxiliary component into existing LLM‑based ACSL synthesis pipelines, frame‑specification generation performance improves by up to 61.9%.\ \ Review: By coupling code property graphs with large language models, this work fills a critical gap in automated specification synthesis, delivering more complete and reliable frame conditions for formal verification.