agentsclimarketplace

Memory deity

Skill v0idOS/performance-deity/archive/memory-deity

A ruthlessly strict algorithmic optimization methodology for AI coding agents. Forces your agent to profile, benchmark, and mathematically prove performance gains before writing code.

Install
npx -y skills add v0idOS/performance-deity --skill memory-deity

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

1.5 KB, as published. Nobody here has run it

Memory Deity: The Leak Hunter

Description

This skill enforces a rigorous approach to tracking down and eliminating memory leaks and excessive garbage collection (GC) pauses. Do not guess where the leak is; prove it with memory profiling.

Triggers

Activate when the user mentions:

  • "memory leak"
  • "high RAM usage"
  • "out of memory" or "OOM"
  • "garbage collection"
  • Or explicitly runs /plugin performance-deity:memory

Core Directives

Phase 1: Isolation

  1. Identify the target code suspected of leaking memory.
  2. Write a continuous execution wrapper that runs the code in an infinite loop or millions of iterations.

Phase 2: Instrumentation

  1. Inject memory tracking into the wrapper:
    • Node.js: Use process.memoryUsage().heapUsed before and after.
    • Python: Use tracemalloc.
    • PowerShell: Use [System.GC]::GetTotalMemory($false).
  2. Run the wrapper and plot the memory delta. If memory continuously grows without dropping after GC, a leak is proven.

Phase 3: Excision

  1. Identify the uncollected references (e.g., event listeners not detached, global arrays growing, closures holding context).
  2. Refactor the code to properly release references, use WeakMaps/WeakRefs, or implement object pooling to reuse memory instead of allocating new objects.

Phase 4: Proof

  1. Re-run the instrumented wrapper.
  2. Present a report showing:
    • Original Memory Growth (e.g., +50MB / 10k ops)
    • New Memory Growth (e.g., +0MB / 10k ops)
    • Explanation of the fix.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.