Generate Reference YAML

Use this skill as the unified backend entrypoint through project CLI.

Do not call IDA API directly in this skill. Always run generate_reference_yaml.py.

Required parameters

• func_name
• gamever (Obtain from dll path via ida-pro-mcp if not specified)
• module (Obtain from dll path via ida-pro-mcp if not specified)
• platform (Obtain from dll path via ida-pro-mcp if not specified)

Command examples

1) Attach to existing MCP (should be used when there is an existing ida-pro-mcp connection)

uv run generate_reference_yaml.py -gamever 14141 -module engine -platform windows -func_name CNetworkGameClient_RecordEntityBandwidth -mcp_host 127.0.0.1 -mcp_port 13337

2) Auto-start idalib-mcp with binary (fallback when no MCP is attached)

# Windows -- always pass -platform windows explicitly
uv run generate_reference_yaml.py -func_name {FUNC_NAME} -auto_start_mcp -binary "bin/{gamever}/{module}/{binary_name}.dll" -platform windows -debug

# Linux -- always pass -platform linux explicitly
uv run generate_reference_yaml.py -func_name {FUNC_NAME} -auto_start_mcp -binary "bin/{gamever}/{module}/lib{module}.so" -platform linux -debug

where {gamever} can be obtained from .env -> CS2VIBE_GAMEVER.

IMPORTANT -- Always pass -platform explicitly. While -platform can theoretically be inferred from the binary extension (.dll -> windows, .so -> linux), auto-inference is unreliable and may produce the wrong platform's reference YAML. Always pass -platform windows or -platform linux explicitly.

3) Custom output name

Use -output_filename when the reference YAML file name should differ from the default <func_name>.<platform>.yaml:

uv run generate_reference_yaml.py -gamever 14141 -module engine -platform windows -func_name CNetworkGameClient_RecordEntityBandwidth -output_filename CNetworkGameClient_RecordEntityBandwidthReference.windows.yaml -mcp_host 127.0.0.1 -mcp_port 13337

-output_filename accepts a file name ending in .yaml, not a path. The file is still written under ida_preprocessor_scripts/references/<module>/. The YAML payload's func_name remains the value passed to -func_name.

IMPORTANT -- Run generate_reference_yaml.py sequentially, NOT in parallel. All invocations share the same IDA MCP connection. Running them in parallel will cause connection conflicts and failures. Run one command at a time, waiting for each to complete before starting the next.

Output path

• Default: ida_preprocessor_scripts/references/<module>/<func_name>.<platform>.yaml
• With -output_filename <name>.yaml: ida_preprocessor_scripts/references/<module>/<name>.yaml

Manual checks after generation

1. func_va is credible for current binary/version.
2. disasm_code is non-empty and matches target function semantics.
3. procedure matches expected semantics when available; it can be an empty string when Hex-Rays is unavailable.
4. func_name only confirms the output file targets your requested canonical name; it does not prove address resolution correctness.

LLM_DECOMPILE path wiring

• Generated file path in repository:
  • ida_preprocessor_scripts/references/<module>/<func_name>.<platform>.yaml
• In target find-*.py, when LLM_DECOMPILE uses relative paths, write:
  • references/<module>/<func_name>.<platform>.yaml
• Example tuple:
  • ("CNetworkMessages_FindNetworkGroup", "prompt/call_llm_decompile.md", "references/engine/CNetworkGameClient_RecordEntityBandwidth.windows.yaml")