Find CEntityResourceManifest AddResource

---
name: find-CEntityResourceManifest_AddResource
description: |
  Locate CEntityResourceManifest_AddResource in CS2 server.dll / libserver.so via IDA Pro MCP and emit a fresh, minimal-unique
  signature or offset for the local gamedata entry "CEntityResourceManifest_AddResource" (symbol CEntityResourceManifest_AddResource). This gamedata entry is stored as an OFFSET. Determine TRUTHFULLY which of the two
  it is: (a) a vtable slot of a POLYMORPHIC class (RTTI/vtable present) -> emit the
  vfunc schema, or (b) a plain struct member of a NON-POLYMORPHIC class (no RTTI/vtable;
  e.g. config-style POD structs like BotProfile) -> emit the structmember schema.
  Never guess: if no vtable exists for the class, it is case (b).
  Trigger: CEntityResourceManifest_AddResource, CEntityResourceManifest_AddResource
disable-model-invocation: true
---

# Find CEntityResourceManifest_AddResource

Target: `CEntityResourceManifest_AddResource` (vfunc) in the CS2 server module.

> Do NOT anchor on raw byte patterns from older releases - they shift. Use anchors only to *locate*
> the function, then generate a fresh minimal-unique function-head signature with relocated bytes
> wildcarded. Produce ONLY the output file(s) listed in this skill's expected outputs, for the binary
> loaded in THIS session (one platform per run). NEVER open or analyze the other platform's binary.

## Method

This gamedata entry is stored as an OFFSET. Determine TRUTHFULLY which of the two
it is: (a) a vtable slot of a POLYMORPHIC class (RTTI/vtable present) -> emit the
vfunc schema, or (b) a plain struct member of a NON-POLYMORPHIC class (no RTTI/vtable;
e.g. config-style POD structs like BotProfile) -> emit the structmember schema.
Never guess: if no vtable exists for the class, it is case (b).

## Mandatory self-check before emitting

The pipeline re-reads the bytes at your `func_va` and deterministically regenerates `func_sig` from
them - if your `func_sig` does not match those bytes EXACTLY (with `??` matching anything), the run
aborts. Therefore:

1. After picking the function, read the actual bytes at `func_va` via the IDA MCP.
2. Derive `func_sig` FROM those bytes: keep stable opcode bytes literally, wildcard relocated or
   variable operands as `??`.
3. `func_sig` MUST start at `func_va` (the true function head).
4. Only then write the YAML. A mismatch is always a bug in YOUR output, never in the pipeline.

## Output schema (STRICT)

Write the YAML file `<symbol>.{platform}.yaml` with EXACTLY these fields:
```yaml
func_name: <SYMBOL_NAME>
func_va: "<hex virtual address>"
func_rva: "<hex rva>"
func_size: "<hex size>"
vtable_name: <owning class RTTI name>
vfunc_offset: "<hex vtable byte offset>"
vfunc_index: <decimal slot index>
```
NEVER include func_sig.

## Verification

1. Decompile the candidate and confirm the behavior matches the purpose above (not a caller or callee).
2. Confirm uniqueness: the generated pattern must match exactly one location in the loaded binary.
3. If a candidate cannot be confirmed, report the shortlist instead of guessing - a skipped symbol is
   safer than a wrong signature.