スキル agent-speedup-estimator
📦
agent-speedup-estimator
コンテンツリビジョン r1 低リスク 🌐 ネットワークアクセス📁 ファイルシステムへのアクセス⚙️ 外部コマンド🔑 環境変数
Measure Agent Speedups on a Faster Computer
Slow local builds, tests, and installs can make coding agents feel unresponsive. This skill separates remote model time from local tool time and estimates hardware-related speedup ranges.
対応: Claude Codex Code(CC)
インストール前に停止して確認を求めてください。
計画を確認し、ファイルを変更する前にユーザーの明示的な同意を得てください。
自分のエージェントでインストール
このリクエストをエージェントにコピーしてください。正規の Skill ページとマニフェストが含まれています。
エージェントリクエスト
Review the Skillstore skill "agent-speedup-estimator" from https://skillstore.io/skills/hnewcity-agent-speedup-estimator.md and its manifest at https://skillstore.io/api/skills/hnewcity-agent-speedup-estimator/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.エージェントは引き続き計画を提示し、セキュリティポリシーで必要な確認を求める必要があります。
エージェントが読めるリソース
AI エージェント、クローラー、スクリプトがページ全体ではなく整理されたコンテキストを必要とする場合は、これらのリンクを使ってください。
テストする
「agent-speedup-estimator」を使用しています。 I want to know whether my laptop slows down Claude Code.
期待される結果:
The report separates remote model time from local tool time, then estimates the speedup available from faster CPU and disk hardware.
「agent-speedup-estimator」を使用しています。 Compare my current computer with an Apple M4 Pro over the last 30 days.
期待される結果:
- Expected gains are limited when most time is spent waiting for the remote model.
- Builds, tests, and installs receive the largest hardware-related benefit.
- The report shows conservative, expected, and optimistic ranges.
「agent-speedup-estimator」を使用しています。 Use my measured CPU scores instead of an inferred baseline.
期待される結果:
The estimate uses the supplied scores, labels the baseline as measured, and preserves uncertainty from workload sensitivity assumptions.
セキュリティ監査
低リスクv1 • 2026/11/1 バージョン別レポートを開く
Confirmed findings are limited to intentional access to hidden Claude Code and Codex transcript directories, which may expose sensitive local history to the skill. The remaining alerts are false positives for static URLs, documentation syntax, report JavaScript, hardware probes, test fixtures, and structured catalog data; no prompt injection or exfiltration intent was found. Static review was capped at 400/3390 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.
25
スキャンされたファイル
31,378
解析済み行数
10
レビュー項目
0
誤検知を無視
確認済みのセキュリティ上の懸念 (1)
低
Unscanned file (too_large) — manual review required
[unscanned: too_large]
Force-confirmed metadata/low static finding; AI false-positive verdict rejected.
機能レビュー項目 (10)
これらは、このスキルに期待される可能性のある実際のローカル機能であるため、レビューが必要ですが、確認済みの悪意ある動作としてはカウントされません。
高
Hidden file in home directory
| Claude Code | `~/.claude/projects/**/*.jsonl` (or `$CLAUDE_CONFIG_DIR`) | message timestamps, `too
The skill reads Claude Code and Codex transcript directories under hidden home paths. This is an intended capability, but those files can contain sensitive prompts, commands, and project data.
高
Hidden file in home directory
| Codex | `~/.codex/sessions` and `archived_sessions` (or `$CODEX_HOME`) | `item_completed` start an
The skill reads Claude Code and Codex transcript directories under hidden home paths. This is an intended capability, but those files can contain sensitive prompts, commands, and project data.
高
Hidden file in home directory
p.add_argument("--claude-dir", help="folder with Claude Code project transcripts (default ~/.claude/
The skill reads Claude Code and Codex transcript directories under hidden home paths. This is an intended capability, but those files can contain sensitive prompts, commands, and project data.
高
Hidden file in home directory
p.add_argument("--codex-dir", help="folder searched recursively for Codex rollout-*.jsonl (default ~
The skill reads Claude Code and Codex transcript directories under hidden home paths. This is an intended capability, but those files can contain sensitive prompts, commands, and project data.
高
Hidden file in home directory
| Transcripts | `~/.claude/projects`, `~/.codex/sessions`, `~/.codex/archived_sessions` (honours `CL
The skill reads Claude Code and Codex transcript directories under hidden home paths. This is an intended capability, but those files can contain sensitive prompts, commands, and project data.
中
Hidden file access
| Claude Code | `~/.claude/projects/**/*.jsonl` (or `$CLAUDE_CONFIG_DIR`) | message timestamps, `too
The referenced paths identify local Claude Code and Codex transcript stores. Access is relevant to the skill, but it creates a real privacy risk because transcript contents may be sensitive.
中
Hidden file access
| Codex | `~/.codex/sessions` and `archived_sessions` (or `$CODEX_HOME`) | `item_completed` start an
The referenced paths identify local Claude Code and Codex transcript stores. Access is relevant to the skill, but it creates a real privacy risk because transcript contents may be sensitive.
中
Hidden file access
p.add_argument("--claude-dir", help="folder with Claude Code project transcripts (default ~/.claude/
The referenced paths identify local Claude Code and Codex transcript stores. Access is relevant to the skill, but it creates a real privacy risk because transcript contents may be sensitive.
中
Hidden file access
p.add_argument("--codex-dir", help="folder searched recursively for Codex rollout-*.jsonl (default ~
The referenced paths identify local Claude Code and Codex transcript stores. Access is relevant to the skill, but it creates a real privacy risk because transcript contents may be sensitive.
中
Hidden file access
| Transcripts | `~/.claude/projects`, `~/.codex/sessions`, `~/.codex/archived_sessions` (honours `CL
The referenced paths identify local Claude Code and Codex transcript stores. Access is relevant to the skill, but it creates a real privacy risk because transcript contents may be sensitive.
リスク要因
🌐 ネットワークアクセス (50)
data/cpu_scores.json:23 data/cpu_scores.json:25 data/cpu_scores.json:26 data/cpu_scores.json:48 data/cpu_scores.json:50 data/cpu_scores.json:51 data/cpu_scores.json:73 data/cpu_scores.json:75 data/cpu_scores.json:76 data/cpu_scores.json:98 data/cpu_scores.json:100 data/cpu_scores.json:101 data/cpu_scores.json:123 data/cpu_scores.json:125 data/cpu_scores.json:126 data/cpu_scores.json:148 data/cpu_scores.json:150 data/cpu_scores.json:151 data/cpu_scores.json:173 data/cpu_scores.json:175 data/cpu_scores.json:176 data/cpu_scores.json:198 data/cpu_scores.json:200 data/cpu_scores.json:201 data/cpu_scores.json:223 data/cpu_scores.json:225 data/cpu_scores.json:226 data/cpu_scores.json:248 data/cpu_scores.json:250 data/cpu_scores.json:251 data/cpu_scores.json:273 data/cpu_scores.json:275 data/cpu_scores.json:276 data/cpu_scores.json:298 data/cpu_scores.json:300 data/cpu_scores.json:301 data/cpu_scores.json:323 data/cpu_scores.json:325 data/cpu_scores.json:326 data/cpu_scores.json:348 data/cpu_scores.json:350 data/cpu_scores.json:351 data/cpu_scores.json:373 data/cpu_scores.json:375 data/cpu_scores.json:376 data/cpu_scores.json:398 data/cpu_scores.json:400 data/cpu_scores.json:401 data/cpu_scores.json:423 data/cpu_scores.json:425
📁 ファイルシステムへのアクセス (23)
README.md:27 README.md:28 README.md:27 README.md:28 scripts/aspeed/cli.py:46 scripts/aspeed/cli.py:47 scripts/aspeed/cli.py:46 scripts/aspeed/cli.py:47 scripts/aspeed/hardware.py:79 scripts/aspeed/hardware.py:87 scripts/aspeed/hardware.py:92 scripts/aspeed/report_html.py:533 SKILL.md:28 SKILL.md:28 tests/test_core.py:5 tests/test_core.py:136 tests/test_core.py:214 tests/test_core.py:233 tests/test_core.py:264 tests/test_core.py:276 tests/test_core.py:285 tests/test_core.py:308 tests/test_core.py:351
⚙️ 外部コマンド (40)
scripts/aspeed/adapters/claude_code.py:20 scripts/aspeed/classify.py:11 scripts/aspeed/classify.py:35 scripts/aspeed/classify.py:94 scripts/aspeed/hardware.py:37 scripts/aspeed/hardware.py:60 scripts/aspeed/report_html.py:202 scripts/aspeed/report_html.py:223 scripts/aspeed/report_html.py:482 scripts/aspeed/report_html.py:483 scripts/aspeed/report_html.py:484 scripts/aspeed/report_html.py:489 scripts/aspeed/report_html.py:490 scripts/aspeed/report_html.py:491 scripts/aspeed/report_html.py:493 scripts/aspeed/report_html.py:496 scripts/aspeed/report_html.py:502 scripts/aspeed/report_html.py:575 SKILL.md:18-20 SKILL.md:20-22 SKILL.md:22 SKILL.md:28 SKILL.md:29 SKILL.md:30 SKILL.md:31 SKILL.md:32 SKILL.md:33-34 SKILL.md:34-45 SKILL.md:45-49 SKILL.md:49-51 SKILL.md:51 SKILL.md:54-55 SKILL.md:55-56 SKILL.md:56-60 SKILL.md:60 SKILL.md:61-62 SKILL.md:62-63 SKILL.md:63 tests/test_core.py:43 tests/test_core.py:249
監査者: codex
このレポートを共有・引用
バージョン付き評価レポート、中立的なバッジ、埋め込みカード、引用を共有できます。Skillstore は証拠を報告しますが、この Skill が安全かどうかは判断しません。
レポートリンクをコピー
https://skillstore.io/skills/hnewcity-agent-speedup-estimator/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdownバッジ
[](https://skillstore.io/skills/hnewcity-agent-speedup-estimator?utm_source=security_passport_badge)HTMLバッジ
<a href="https://skillstore.io/skills/hnewcity-agent-speedup-estimator?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/hnewcity-agent-speedup-estimator/security.svg" alt="Skillstore security assessment" loading="lazy"></a>埋め込みカード
<iframe src="https://skillstore.io/embed/skills/hnewcity-agent-speedup-estimator.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>学術引用 (APA · BibTeX · CFF)
APA形式の引用
hnewcity. (2026). agent-speedup-estimator security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/hnewcity-agent-speedup-estimator/audits/1BibTeX形式の引用
@techreport{hnewcity-hnewcity-agent-speedup-estimator-2026,
author = {hnewcity},
title = {agent-speedup-estimator security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/hnewcity-agent-speedup-estimator/audits/1},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "agent-speedup-estimator security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
- name: "hnewcity"
date-released: "2026-11-01"
url: "https://skillstore.io/skills/hnewcity-agent-speedup-estimator/audits/1"
identifiers:
- type: other
value: "skillstore:hnewcity-agent-speedup-estimator:audit:1"
description: "Skillstore immutable audit report identifier"
Skillstore スコア
このスコアの理由 証拠の信頼度: 中 59
アーキテクチャ
85
保守性
87
コンテンツ
65
コミュニティ
91
仕様準拠
作成できるもの
Audit a personal workstation
Measure how much of your agent work is local and estimate whether a faster computer could reduce waiting time.
Plan a team hardware refresh
Use aggregate timing reports to compare likely gains before standardizing a new developer machine.
Compare a named target machine
Evaluate a specific laptop or desktop against the current system with documented CPU score assumptions.
これらのプロンプトを試す
Get a baseline report
Analyze my available Claude Code and Codex history. Separate model time from local tool time and summarize the main sources of delay.
Compare two target machines
Compare my current computer with [target one] and [target two]. Show expected speedup ranges, hours saved, and the assumptions behind each estimate.
Review a recent period
Analyze the last [number] days of agent activity. Focus on builds, tests, installs, and scripts, then explain which local categories benefit most from a hardware upgrade.
Calibrate with measured scores
Use my measured Geekbench 6 single-core score of [single] and multi-core score of [multi] as the current baseline. Compare it with [exact target] and state all uncertainty sources.
ベストプラクティス
- Name an exact target CPU or machine when comparing hardware.
- Read the assumptions and ranges before treating a result as a purchase decision.
- Keep transcript directories and generated reports private because local history can contain sensitive data.
回避
- Do not treat a CPU score comparison as a benchmark of your complete workload.
- Do not share raw transcripts or reports without reviewing activity dates and hardware details.
- Do not compare Geekbench scores from different major versions as if they were equivalent.
よくある質問
Does this measure model response speed?
It estimates model time from local timing records, but remote model inference is treated as unchanged by your computer.
What data does the skill read?
It reads local Claude Code and Codex timing records, hardware metadata, and bundled CPU reference data.
Are prompts and replies included in reports?
The skill is designed to exclude prompts, replies, raw commands, transcript paths, and session identifiers from reports.
Can I compare a specific laptop or desktop?
Yes. Use an exact catalog name when available, and review whether the estimate uses measured scores, a stand-in, or a fallback.
Are the speedup numbers guaranteed?
No. They are scenario ranges based on CPU scores and sensitivity assumptions, not guarantees or full workload benchmarks.
Does it support both Claude Code and Codex?
Yes. It has adapters for both tools and reports where usable timing records are available.