octocode-benchmark

octocode-benchmark is a skill for Claude Code, Codex from bgauryy/octocode. It costs 117 tokens per session (2,215 once invoked), scanned A, original, MIT.

A procedure for comparing Octocode with other code-research setups by having separate agents answer the same questions and a blind judge compare their answers.

In plain words
What is it for?
Use it to run benchmark rounds, validate the runs, collect answers and judging logs, and produce paired comparison statistics and reports.
Why use it?
It provides a repeatable way to measure answer quality and transcript length across competing setups.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is OCTO_LOG="$CAMP/octocode-p1-Q4.jsonl" ./compare/bin/octoc1822 ghGetFileContent \.

Good fit Use it to run benchmark rounds, validate the runs, collect answers and judging logs, and produce paired comparison statistics and reports.

Compare 6 skills from other repositories ↓
Install

Getting it into your agent

It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.

Clone the repo
git clone --depth 1 https://github.com/bgauryy/octocode
agentmods
npx agentmods add skills/bgauryy/octocode/octocode-benchmark

Made for: Claude Code, Codex.

Wrote 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.

agentmods badge for octocode-benchmark

README.md
[![agentmods](https://agentmods.dev/badge/skills/bgauryy/octocode/octocode-benchmark/github.svg)](https://agentmods.dev/skills/bgauryy/octocode/octocode-benchmark)
Your own site
<a href="https://agentmods.dev/skills/bgauryy/octocode/octocode-benchmark"><img src="https://agentmods.dev/badge/skills/bgauryy/octocode/octocode-benchmark/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.

agentmods 80×15 button for octocode-benchmark

Your own site · 80×15
<a href="https://agentmods.dev/skills/bgauryy/octocode/octocode-benchmark"><img src="https://agentmods.dev/badge/skills/bgauryy/octocode/octocode-benchmark.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 117 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,215 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 2 findings, up to medium

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • medium MCP Rug Pull · line 38
    npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.
    Fix: Pin the version: npx @scope/[email protected]
  • medium MCP Rug Pull · line 48
    npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.
    Fix: Pin the version: npx @scope/[email protected]
How audits are shown
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 5.1 $0.00117 $0.02215
Opus 5 $0.00059 $0.01107
Sonnet 5 $0.00023 $0.00443
Haiku 4.5 $0.00012 $0.00221

Measured 11d ago against content hash cef37ebcb5fd, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

octocode-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 11d ago.

The scan reads SKILL.md. This mod also ships 4 executable files (scripts/_hr_check.py, scripts/_hr_import.py, scripts/check-prereqs.sh, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

packages/octocode-benchmark/skills/octocode-benchmark/SKILL.md · 134 lines

How it starts

The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Octocode benchmark

Plain-markdown, run-by-hand CLI research comparison. Octocode is the anchor; each baseline is a separate pairwise matchup (octocode vs rtk | headroom | gh). Per question, per pass: two isolated runners answer, one blind judge grades them X / Y (randomized per question). Run ≥3 passes; the rollup shows every matchup together. No harness, no JSON.

Paths below are relative to the package root packages/octocode-benchmark/. Shared tooling lives in compare/bin/; questions in compare/github-questions/; reports in results/.

Flow (4 phases)

  1. Preflight — verify + pin every arm; a failure invalidates the run.
  2. Answer — 2 isolated runners (anchor + baseline) per question/pass, leanest-legal path, each appends a ## Q<n> section to answers/<arm>-p<pass>.md.
  3. Judge — after both sections exist, one blind judge reasons to a verdict, then scores.
  4. Summarize — validate logs, aggregate paired stats, update the rollup.

How it is MEASURED (characters, never self-reported)

The metric is total_chars = model-in + model-out in Unicode code points, from an instrumented log — the tool transcript only (excludes system prompt, tool schemas, model reasoning; the fixed per-arm primer is excluded by rule; any later help/schema/failed call is counted).

  • model-out = the command string + args the model wrote, plus the final answer.
  • model-in = the tool output pulled back into context (for Headroom, the compressed output).

Mechanism: every research command runs through its arm's thin wrapper, which shells the real CLI unchanged, prints output verbatim, and appends one JSONL row per call:

Arm Wrapper Runs Log env
octocode (local build) compare/bin/octoc npx octocode tools … OCTO_LOG
octocode (published pin) compare/bin/octoc1822 npx -y [email protected] tools … OCTO_LOG
gh+RTK compare/bin/rtkm rtk gh … RTK_LOG
gh+Headroom compare/bin/ghc gh … → Headroom compress GHC_LOG (+ HR_PY)
plain gh compare/bin/ghm gh … (read-only)

Read the full file on GitHub · 134 lines

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. 11d ago First seen · 134 lines · 117 tokens per session scan A cef37ebcb5fd

Subscribe to this mod's changes

octocode-benchmark is a skill published in the GitHub repository bgauryy/octocode (932 stars, last pushed yesterday), licensed MIT. It adds 117 tokens to every session and 2,215 once invoked, about $0.0006 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.