SkillsBench is a benchmark for measuring how effectively AI agents use modular skills—folders containing instructions, scripts, and resources—to complete specialized tasks. It helps researchers and developers evaluate both skill quality and agent behavior, including tasks that require combining multiple skills. The catalogue’s skills and instructions are evaluated as part of this workflow.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add benchflow-ai/skillsbench --skill cache-policy-comparisongit clone --depth 1 https://github.com/benchflow-ai/skillsbenchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/benchflow-ai/skillsbench/cache-policy-comparison)<a href="https://agentmods.dev/skills/benchflow-ai/skillsbench/cache-policy-comparison"><img src="https://agentmods.dev/badge/skills/benchflow-ai/skillsbench/cache-policy-comparison/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/benchflow-ai/skillsbench/cache-policy-comparison"><img src="https://agentmods.dev/badge/skills/benchflow-ai/skillsbench/cache-policy-comparison.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00102 | $0.01272 |
| Opus 5 | $0.00051 | $0.00636 |
| Sonnet 5 | $0.00020 | $0.00254 |
| Haiku 4.5 | $0.00010 | $0.00127 |
Grade A, and why
cache-policy-comparison scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 86 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Overview
An eviction policy decides which resident entry a cache removes when a new entry is admitted beyond capacity. Four policies cover almost every replay-and-measure task:
| Policy | Data structure | On hit | On admit | Eviction choice |
|---|---|---|---|---|
| LRU | OrderedDict | Move to tail | Append at tail | Pop head |
| LFU | {key: freq} + insertion order |
freq[k] += 1 |
freq[k] = 1 |
Min freq, tiebreak by insertion order |
| FIFO | OrderedDict | Nothing | Append at tail | Pop head |
| S3FIFO | Three FIFO queues + freq[k] |
freq[k] = min(freq+1, cap) |
Admit to small; ghost-hit admits to main | Second-chance on main; small drains to main/ghost |
Each has subtleties that trip naive implementations.
LRU
Use an OrderedDict where the tail is the most-recently-accessed key. On hit, move_to_end. On miss + insert, append; pop from head if over capacity.
Most common bug: forgetting to update recency on a hit. Without the refresh, LRU degenerates to FIFO — hit rate drops substantially on any workload with recency structure.
from collections import OrderedDict
class LRU:
def __init__(self, capacity):
self.capacity = capacity
self._d = OrderedDict()
def contains(self, k): return k in self._d
def access(self, k):
if k in self._d:
self._d.move_to_end(k)
else:
self._d[k] = None
if len(self._d) > self.capacity:
self._d.popitem(last=False)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 86 lines · 102 tokens per session scan A 353a72157276
cache-policy-comparison is a skill published in the GitHub repository benchflow-ai/skillsbench (1,764 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 102 tokens to every session and 1,272 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
rag-eval
NVIDIA RAG Blueprint evaluation guidance for measuring retrieval and answer quality with stable datasets, baselines, and reproducible scoring workflows.
local-llm-ops
Local LLM operations with Ollama on Apple Silicon, including setup, model pulls, chat launchers, benchmarks, and diagnostics.
fixed-tensor-testing
Test ML functions with fixed input tensors for reproducibility.
dbscan-custom-metric
Run DBSCAN clustering with a custom distance metric using sklearn, including how to define weighted Euclidean metrics and extract cluster centroids.
parallel-grid-search
Parallelize hyperparameter grid search using joblib for efficient multi-core execution.
json-data-extraction
Extract, parse, and query JSON data from large enterprise files efficiently.