<< All versions

Skill v1.0.0

currentLLM-judged scan90/100
1llu50ry/mcp-gamehacking/frame-observation-boundary
──Details
PublishedOctober 24, 2026 at 11:29 AM
Content Hashsha256:2e9f0d7541bf28bf...
Git SHAd2b7d4946381
──Files
Files (1 file, 3.8 KB)
SKILL.md3.8 KBactive
SKILL.md · 67 lines · 3.8 KB

version: "1.0.0" name: frame-observation-boundary description: "Discipline for interpreting graphics, overlay, and screenshot evidence: record the observation layer and baseline dimensions before naming a hook, cheat, or anti-cheat mechanism. Pair with [[resea" metadata: type: game-security source: awesome-game-security/wiki topics: [graphics-api, anti-cheat, game-hacking]


Frame Observation Boundary

Discipline for interpreting graphics, overlay, and screenshot evidence: record the observation layer and baseline dimensions before naming a hook, cheat, or anti-cheat mechanism. Pair with [[research-rigor]] when a README sample or capture artifact is treated as proof.

Baseline before interpretation

Record for every rendering or capture claim:

DimensionWhy it matters
Graphics API and backendDX9–12, OpenGL, Vulkan, translation layers (DXVK, Wine) change hook points
Driver and compositorDWM, flip model, DirectFlip, Independent Flip, HDR/SDR alter what a capture sees
Capture pathGame inject, Desktop Duplication, PrintWindow, Present hook, GPU readback, NVIDIA scanout
Synchronization modelFrame discard, vsync, staging/readback timing affect stale or black frames
Tool and versionOBS mode, AC module, profiler, or debugger version changes artifacts

Black, missing, or stale frames have multiple benign explanations—they are not sufficient evidence of concealment without a controlled comparison.

Layer separation

Do not treat these as interchangeable copies of the same image:

Application render targets
→ presentation queues / swap-chain Present
→ compositor output (DWM, flip model)
→ physical display output
→ captured image (screenshot, OBS, AI pipeline)

Each step can drop, reorder, or transform content under documented platform behavior. Route hook-specific questions to [[present-hook]] or [[draw-call-hook]]; capture-contract questions to [[anti-screenshot-capture]] or [[obs-game-capture]].

Capability classification

Classify claims by required capability before attribution:

ClassExampleEvidence focus
In-process graphics changePresent hook, draw-call chams, shader injectModule provenance, vtable/state integrity, frame debugger
Separate-process overlayLayered HWND, DWM draw, Steam/NVIDIA hijackWindow affinity, composition surface, ETW/GDI signals
Frame access for analysisOBS Game Capture, DDA, AC screenshotActive capture API, inject footprint, readback bandwidth

Include counterexamples: legitimate streaming (OBS), accessibility overlays, vendor tools (GeForce Experience, Steam), debug visualization, and post-processing injectors such as [[reshade]].

Capture vs enforcement

Observed capture behavior (a PNG, OBS hook DLL, staging texture pattern) is collection evidence. Assumed anti-cheat enforcement (ban, flag, staff review) requires the detector's configured path, policy, and corroboration—do not infer enforcement scope from a single screenshot or hook artifact alone. Scheduled snapshots cover their acquisition intervals; intermittent absence does not establish session-wide absence.

Coverage matrix (owned samples)

For owned applications or lab repros, document: API, OS/driver build, windowed vs exclusive fullscreen, presentation model, monitor count, capture backend, cursor handling, and timestamps. Use known-content controls to distinguish capture failure, protected content, output selection, and actual rendering differences.

Related

[[present-hook]] · [[draw-call-hook]] · [[anti-screenshot-capture]] · [[obs-game-capture]] · [[world-to-screen]] · [[cheat-attack-surface]] · [[research-rigor]] · [[overviews/graphics-api]] · [[overviews/game-hacking]] · [[overviews/anti-cheat]]

All versions