Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Shader Optimization Skill · a5c-ai | DecimalAI

a5c-ai/shader-optimization

NewSkillScoreMaintained

Shader performance optimization skill for instruction counting, GPU profiling, and rendering efficiency.

codingby a5c-ai

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Coding pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/a5c-ai-shader-optimization/SKILL.md

decimalai skills pull a5c-ai-shader-optimization --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/a5c-ai-shader-optimization/SKILL.md https://app.decimal.ai/s/a5c-ai-shader-optimization/SKILL.md

Raw file: /s/a5c-ai-shader-optimization/SKILL.md ↗ · machine-readable summary at /s/a5c-ai-shader-optimization.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 27d ago

No measurable change+1% tokens0% turns91% of whole cases pass with skill (20/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-03	✓→✓	＝ Same ✓	-8%	0%
case-04	✓→✓	＝ Same ✓	-4%	0%
case-05	✓→✓	＝ Same ✓	-4%	0%
case-06	✓→✓	＝ Same ✓	-7%	0%
case-07	✓→✓	＝ Same ✓	-14%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

a5c-ai / babysitter

library/specializations/game-development/skills/shader-optimization/SKILL.md

View on GitHub ↗

License MITImported 8/10/2026Version v1

Skill contents

SKILL.md

Shader Optimization Skill

Shader performance optimization techniques.

Overview

This skill provides capabilities for optimizing shader performance.

Capabilities

Instruction counting
ALU optimization
Texture sample reduction
Shader variant management
GPU profiling

Best Practices

Minimize texture samples
Use half precision where possible
Avoid branching
Batch similar shaders
Profile on target hardware

References

Engine-specific profiling tools

Illustrative Want to see what a5c-ai/shader-optimization would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/29/2026, 14:38 UTC · 22 cases

Lift

0%

91% vs 91%

Agent turns

0%

1.0 vs 1.0

Tokens

+1%

2,419.59 vs 2,388.59

pass → pass · 20fail → fail · 2

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 Our GPU profiler shows high latency in our terrain surface shader due to heavy math and multiple sampler calls. Please analyze the shader code I've provided and output a detailed optimization plan along with rewritten code snippets.	fail→fail	13,655	28,363	+108%	1	1	0%	2,160	4,218	+95%	0	0	—
case-02 We have multiple shader variants causing pipeline stalls in our graphics engine. Could you audit our GLSL code structure and return a prioritized list of refactoring steps to streamline execution across target hardware?	fail→fail	13,147	21,869	+66%	1	1	0%	2,114	3,239	+53%	0	0	—
case-03 We are writing a custom HLSL specular lighting calculation for mobile GPUs where vector math is causing bottlenecks. We are currently using float4 for all vectors. What data type optimization should we apply?	pass→pass	14,902	13,370	-10%	1	1	0%	2,143	1,969	-8%	0	0	—
case-04 Our material shader fetches separate albedo, normal, roughness, and metalness textures in distinct sampler calls. What optimization strategy reduces the sampler overhead on the GPU?	pass→pass	14,820	12,017	-19%	1	1	0%	2,130	2,055	-4%	0	0	—
case-05 We have conditional logic in our pixel shader `if (height > threshold) { color = vec4(1.0); } else { color = vec4(0.0); }`. How should this logic be rewritten for optimal GPU ALU execution?	pass→pass	12,486	9,357	-25%	1	1	0%	1,765	1,686	-4%	0	0	—
case-06 Our rendering engine performs draw calls with tiny variations in material parameters by binding distinct shader objects for each object. How can we optimize draw execution at the engine level?	pass→pass	16,466	15,129	-8%	1	1	0%	2,670	2,480	-7%	0	0	—
case-07 Before shipping our mobile graphics pipeline, how should we validate that our ALU optimizations actually yielded framerate improvements?	pass→pass	18,384	15,375	-16%	1	1	0%	2,607	2,230	-14%	0	0	—
case-08 We want to measure the mathematical instruction density of a post-processing fragment shader before running it. What metric analysis step should be performed?	pass→pass	11,978	15,240	+27%	1	1	0%	1,949	1,649	-15%	0	0	—
case-09 In HLSL, we have separate multiply and add operations: `float result = (a * b) + c;`. How should math operations be structured for optimal ALU throughput?	pass→pass	16,663	18,254	+10%	1	1	0%	2,767	2,702	-2%	0	0	—
case-10 Our build system generates over 10,000 permutation variants for a single surface shader. What pipeline capability should be applied to manage these permutations?	pass→pass	13,305	9,767	-27%	1	1	0%	1,989	1,337	-33%	0	0	—
case-11 When optimizing texture coordinate calculations for large tiled landscapes in fragment shaders, should UV coordinates be cast to half precision?	pass→pass	18,560	16,805	-9%	1	1	0%	2,508	2,299	-8%	0	0	—
case-12 Our distant foliage shader is suffering from high texture bandwith overhead. Beyond packing channels, what sampling practice reduces bandwidth for distant surface fragments?	pass→pass	19,776	14,563	-26%	1	1	0%	2,510	2,150	-14%	0	0	—
case-13 We use `clip()` or `discard` in our alpha-tested foliage shader, but frame rates dropped on tile-based deferred GPUs. Why does this conditional flow impair performance?	pass→pass	16,599	18,261	+10%	1	1	0%	2,490	2,706	+9%	0	0	—
case-14 We need to identify whether our rendering bottleneck is vertex fetch bound or fragment ALU bound on an iOS device. What step should be taken?	pass→pass	11,359	9,297	-18%	1	1	0%	1,637	1,499	-8%	0	0	—
case-15 We are calculating expensive trigonometric functions (`sin`, `cos`) per-pixel versus reading a precomputed 1D lookup texture. Which approach is generally preferred for modern shader performance?	pass→pass	15,277	16,752	+10%	1	1	0%	1,976	2,179	+10%	0	0	—
case-16 Our HLSL shader uniforms are declared as individual float variables scattered across memory. How should uniform data structures be organized for efficient GPU access?	pass→pass	19,712	15,821	-20%	1	1	0%	2,841	2,693	-5%	0	0	—
case-17 We have a fixed loop iterating 4 times in a blur shader. Should the loop remain as a dynamic execution loop or be unrolled?	pass→pass	12,040	11,906	-1%	1	1	0%	1,804	1,657	-8%	0	0	—
case-18 We calculate light attenuation per fragment in the pixel shader. The result varies smoothly across the mesh. How can ALU workload be optimized?	pass→pass	15,008	14,766	-2%	1	1	0%	2,369	2,468	+4%	0	0	—
case-19 We need to reduce the high polygon count of our FBX character models in Maya before importing them into our engine. What mesh processing steps should we follow?	pass→pass	17,769	16,080	-10%	1	1	0%	2,824	2,335	-17%	0	0	—
case-20 Our C++ game engine main thread is blocked while preparing render command lists. How can we refactor our C++ task graph to utilize multi-core CPUs?	pass→pass	35,091	20,995	-40%	1	1	0%	3,529	3,879	+10%	0	0	—
case-21 We are implementing a low-pass audio filter in C# for our game's sound engine. What algorithm should we use to smooth high-frequency sample buffers?	pass→pass	21,530	17,205	-20%	1	1	0%	3,301	3,123	-5%	0	0	—
case-22 Our game installer executable size is too large for distribution. How should we configure Zstandard compression settings for our asset pack files on disk?	pass→pass	18,291	16,872	-8%	1	1	0%	2,466	2,678	+9%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of 0 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 90.9%
Pass rate with the skill: 90.9%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/29/2026, 14:38 UTC

SkillScore blends 4 signalsno signals yet

BenchmarkRun a verified benchmark with positive lift

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns-5% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

No signals yet — run a benchmark or collect evaluated traces to earn a score.

Verified on gemini-3.6-flash · re-tested 26 days ago

Other measured skills in the registry, with their headline benchmark lift.