pts-pivotal-token-search
Analyze Local Model Reasoning with PTS
Reasoning traces do not show which events change a model's chance of solving a task. This skill guides PTS searches, event interpretation, and local dataset exports with explicit measurement caveats.
Установить с помощью моего Агента
Скопируйте этот запрос в своего Агента. Он содержит каноническую страницу Skill и манифест.
Review the Skillstore skill "pts-pivotal-token-search" from https://skillstore.io/skills/lambdasec-pts-pivotal-token-search.md and its manifest at https://skillstore.io/api/skills/lambdasec-pts-pivotal-token-search/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.Ваш Агент по-прежнему должен показать план и запросить все подтверждения, требуемые политикой безопасности.
Ресурсы для AI-агентов
Используйте эти ссылки, когда AI-агенту, crawler или script нужен чистый контекст вместо полной страницы.
Протестировать
Использование «pts-pivotal-token-search». Explain how to compare token events and latent events in my PTS file.
Ожидаемый результат:
Rank token events by absolute probability change and retain the direction of the shift. Report latent readout scores separately. Do not combine these scales into one ranking.
Использование «pts-pivotal-token-search». Plan my first token search without starting it.
Ожидаемый результат:
Confirm the model, dataset, device, and time budget. Inspect the installed CLI flags. Propose fewer examples and samples for an initial pass. Request download consent and disclose the expected cost before execution.
Использование «pts-pivotal-token-search». Prepare existing events for a local DPO export.
Ожидаемый результат:
Link the local events, inspect available export filters, and test DPO export on a small file. Review the sample before scaling up. Do not publish or upload the results.
Аудит безопасности
БезопасноAll 30 static findings are false positives: Markdown backticks, legitimate local PTS workflows, and relevant project links. The skill requires consent for downloads and publishing, prohibits automatic installation, and cautions against unsupported causal claims. No evidence found of prompt injection or unauthorized data transfer; external PTS implementations and hosted services were not audited.
Факторы риска
⚙️ Внешние команды (27)
🌐 Доступ к сети (3)
Поделиться и цитировать этот отчет
Делитесь версионным отчетом об оценке, нейтральным значком, встраиваемой карточкой и цитатами. Skillstore публикует доказательства, не решая, безопасен ли этот Skill.
Копировать ссылку на отчёт
https://skillstore.io/skills/lambdasec-pts-pivotal-token-search/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportЗначок Markdown
[](https://skillstore.io/skills/lambdasec-pts-pivotal-token-search?utm_source=security_passport_badge)Значок HTML
<a href="https://skillstore.io/skills/lambdasec-pts-pivotal-token-search?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/lambdasec-pts-pivotal-token-search/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Встраиваемая карточка
<iframe src="https://skillstore.io/embed/skills/lambdasec-pts-pivotal-token-search.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>Академические ссылки (APA · BibTeX · CFF)
Цитата APA
lambdasec. (2026). pts-pivotal-token-search security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/lambdasec-pts-pivotal-token-search/audits/1Цитата BibTeX
@techreport{lambdasec-lambdasec-pts-pivotal-token-search-2026,
author = {lambdasec},
title = {pts-pivotal-token-search security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/lambdasec-pts-pivotal-token-search/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: "pts-pivotal-token-search security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
- name: "lambdasec"
date-released: "2026-10-28"
url: "https://skillstore.io/skills/lambdasec-pts-pivotal-token-search/audits/1"
identifiers:
- type: other
value: "skillstore:lambdasec-pts-pivotal-token-search:audit:1"
description: "Skillstore immutable audit report identifier"
Оценка Skillstore
Почему такая оценка Достоверность доказательств: СреднийЧто вы можете построить
Investigate pivotal reasoning events
Plan a small token or sentence search to identify events associated with changes in task success.
Inspect a local evaluation trace
Review existing event files, compare events within one scale, and identify examples that need manual checking.
Prepare a local training dataset
Link existing events and test a filtered DPO export before processing a larger collection.
Попробуйте эти промпты
Explain the fields in my local PTS events file at [path]. Summarize its scales and model identifiers without running searches or uploading data.
Plan a token search for [model] on [dataset] using [device] within [budget]. Check available flags and ask before downloads or execution.
Review [events path] and plan J-lens enrichment for [model]. Include shuffled controls, consent for downloads, and separate reporting of latent scores.
Design repeated-seed checks and an intervention for [events path], then propose a filtered local DPO export. Separate observational findings from causal evidence.
Лучшие практики
- Confirm the model, dataset, device, budget, and download permissions before starting a small pilot search.
- Compare events within one scale and model, inspect examples manually, and repeat measurements with other seeds.
- Use shuffled controls for latent enrichment and require interventions before making causal claims.
Избегать
- Do not combine latent readout scores and emitted-event probability shifts into one ranking.
- Do not treat heuristic event links or category labels as verified causal mechanisms.
- Do not install dependencies, download assets, or publish private events without the required user consent.