benchmark

A repeatable benchmark for measuring Sipcode's token savings on a fixed set of 20 coding tasks.

In plain words
What is it for?
It is for testing Sipcode with the full benchmark or a smaller three-task smoke test and reporting the measured result.
Why use it?
It gives users a way to check the claimed savings on their own machine rather than relying only on an advertised result.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/anuj7411/sipcode/benchmark
Any agent
npx skills add Anuj7411/sipcode --skill benchmark
Clone the repo
git clone --depth 1 https://github.com/Anuj7411/sipcode

Made for: Claude Code, Codex.

Per session 68 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 314 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00068 $0.00314
Opus 5 $0.00034 $0.00157
Sonnet 5 $0.00014 $0.00063
Haiku 4.5 $0.00007 $0.00031

Measured 2d ago against content hash b8ede9dbe9ad, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

benchmark 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 2d 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.

skills/benchmark/SKILL.md · 20 lines

What it actually says

Sipcode — Benchmark (the reproducibility proof)

Tell the user to run npx sipcode benchmark in their terminal (this is a CLI command, not an MCP tool — the benchmark requires file I/O outside the MCP scope). It takes ~90 seconds.

If they want a quick smoke (3 tasks, ~15s), tell them to run npx sipcode benchmark --quick.

When the user reports back the output:

  • The headline is the median savings % across the 20 tasks. Target: ~62.6%, range typically 37.4% – 80.6%.
  • Surface where the savings came from — typically:
    • ~30% from S001 smart manifest
    • ~34% from S021 output compression
    • ~36% from S030 read-once cache
  • Be explicit that these are simulation numbers computed against the locked corpus, not a live Claude session A/B. The methodology is at benchmark/METHODOLOGY.md for reproducibility.

If the user's measured savings differ significantly from the 62.6% median, that's interesting — surface it as a real workload signal, not a bug.

Changes

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.

  1. 2d ago First seen · 20 lines · 68 tokens per session scan A b8ede9dbe9ad

Subscribe to this mod's changes

benchmark is a skill published in the GitHub repository Anuj7411/sipcode (46 stars, last pushed 1mo ago), licensed MIT. It adds 68 tokens to every session and 314 once invoked, about $0.0003 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 2026-08-30.

Related

Other skills, from other repositories

audit-pr-architecture

Audit a PR's structural choices against architecture-priming, the repo's own architecture.md, and any nested boundary configs. Use when the user says "audit this PR's architecture", "check the boundaries", "is this PR clean structurally", "post-merge review of.

stainless-code/codemap · 61 tokens

docs-voice

Voice, tone, and format for the public Codemap docs (apps/docs, built with Blume). Use when authoring or editing apps/docs prose — landing, guides, concepts, recipes, reference, generated API — or deciding headline grammar, benefit framing, peer framing, or anti-pitch wording.

stainless-code/codemap · 65 tokens

improve-codebase-architecture

Find deepening opportunities in the codebase, informed by the domain language in docs/glossary.md and the architecture in docs/architecture.md. Use when the user wants to improve architecture, find refactoring opportunities, consolidate tightly-coupled modules, or make a codebase more testable and AI-navigable.

stainless-code/codemap · 71 tokens

writing-agents-config

Codemap repo hybrid deltas for .agents/ — tiers, rule-skill pairing, repo exemplars. Use when creating or reviewing rules/skills in this repo. Read writing-great-skills first for skill vocabulary and authoring principles.

stainless-code/codemap · 54 tokens

agents-tier-system

Full tier assignments, pairing conventions, and authoring checklist for .agents/ rules and skills. Use when creating or reviewing a rule or skill, deciding Tier 1 vs 2 vs 3, or auditing attachment cost.

stainless-code/codemap · 49 tokens

verify-after-each-step

Per-file verification checklist — lint-staged and package.json scripts after each milestone.

stainless-code/codemap · 22 tokens