Перейти к содержимому
New

DeepSeek V4.1 Flash free for 7 days: 10M tokens a day after a top-up of $10 or more. Until Nov 30.

How it works
aeondave/loop-control-and-pivotsСети, беспроводная связь и телеком

loop-control-and-pivots

⁨Stop unproductive retries and repeated polling. Use when the same failure recurs, a side issue consumes the task, or unfinished jobs are checked without new evidence.⁩

v2026.11.13-21cd5683 установки+3 за 7 дней39 ЗвездыMITсинхронизировано 10 часов назад

Установка

Ещё нет NotHoly CLI? Сначала установите её.

$ curl -fsSL https://notholy.com/install.sh | sh
$ notholy skills install aeondave/loop-control-and-pivots

Пишет в ⁦~/.notholy/skills⁩ после проверки подписи и SHA-256.

ПодписьПодписано, ed25519
sTDvyA+bqu…BrqH+4Aw==CLI проверяет SHA-256 и подпись, прежде чем записать хоть один файл.
Пакетc6cc7a72…1c1e7ffb
c6cc7a729295…b7681c1e7ffb1 файл · ⁦2.0 KB⁩
ПроверкаНизкийзамечаний нет
Проверка не нашла в этой версии ничего подозрительного.Вся проверка
Источникgithub.com/AeonDave/malskill/tree/main/behaviours/loop-control-and-pivots@21cd568

Эта запись еще не переведена на русский, поэтому показан текст на другом языке: English.

Keep the requested outcome while changing approaches that no longer yield evidence.

Retry or change course

  • After repeated equivalent failures, compare the expected and actual result before another attempt. A small retry count is a reassessment cue, not proof that a cause or solution is impossible.
  • Do not repeat unchanged authentication, permission, malformed-input, or unsupported-feature failures. Correct a demonstrated cause, use a supported route, or identify the missing capability.
  • Retry transient failures only within the tool's retry policy and task budget, respecting backoff or server guidance.
  • Record the failed approach, decisive error, and what the next attempt changes. Preserve useful partial results instead of restarting discovery.
  • Give side issues a budget proportionate to their role in the deliverable. Continue independent work when a dependency is unresolved.

Use hypothesis-driven when the causal model is uncertain and known-problem-hint-research when external evidence could clarify an unfamiliar problem or reveal the next useful test. Do not invent additional work simply to avoid reporting a limitation.

Persistence and stopping

A failed approach does not automatically block the whole task. Check remaining supported alternatives that fit the authorization and budget. Stop at an explicit limit, an unavailable required capability, or a request to pause.

Distinguish a proven constraint from an untested assumption. Static evidence or a valid proof can justify a bounded conclusion; unsuccessful sampling alone cannot prove impossibility. Do not demand live actions outside the authorized scope to support every negative claim.

Waiting for running work

  • For a long-running job, identify its completion signal and execution deadline. Check the current host's tool contract before choosing a wait or notification mechanism.
  • Prefer supported completion notifications or native waits. Do independent useful work meanwhile; when exhausted, use a wait suited to the expected duration within tool limits and higher-priority responsiveness requirements.
  • If polling is the only supported option, space checks with backoff bounded by the job deadline and required intervention latency. Avoid short sleep/check loops and rereading unchanged logs.
  • A wait returning early or timing out does not establish job failure. Check its reported state; silence alone does not justify cancellation, relaunch, or counting a failed attempt.
  • End a turn with work pending only if the job can survive it; also require reliable host resumption or a user-agreed handoff. State the job, result location, and resume mechanism; otherwise retain responsibility for completion. Instructions alone do not create wake-ups.
  • On completion, verify exit status and required artifacts before continuing dependent work. A delivered notification is not evidence of success.

Budgets and handoff

Respect explicit runtime, cost, retry, and scope limits. If estimating a budget, scale it to the uncertainty and retain enough capacity for verification and reporting; no universal call count proves exhaustion.

For delegated work, enforce the agreed budget and investigate a missed milestone, reported failure, or scope drift before reassignment. Apply the running-work rules to a quiet worker.

When blocked, report the missing capability, exact failure or constraint, useful results, and the smallest action that would unblock the work. Follow the host's task-state contract before marking a goal blocked; the label is not a substitute for evidence.

Файлы этой версии

Права

ПравоЗаявленоНайдено проверкой
Читает файлыНетНет
Пишет файлыНетНет
Выполняет команды shellНетНет
Ходит в сетьНетНет
Использует коннекторыНетНет
Управляет браузеромНетНет

Замечания проверки

Проверено 13 нояб. 2026 г.

Проверка не нашла в этой версии ничего подозрительного.

Чистая проверка снижает риск, но не доказывает, что пакет безопасен. Читайте то, что ставите.

ВерсияОпубликованаКоммитSHA-256ПодписьИзменение прав
v2026.11.13-21cd56813 нояб. 2026 г.21cd568c6cc7a72Подписана–

Спрашивайте обо всём, что позволяет ваш скоуп.

Создайте аккаунт, пополните баланс и задайте первый вопрос в чате или в терминале.

  1. 1/3

    Создайте аккаунт

    Личный или для компании. Позже можно перейти на аккаунт компании без потери баланса и ключей.

    ЛичныйКомпания
  2. 2/3

    Пополните баланс

    Картой. Оплата за токены, без подписки.

    $5.00минимальное пополнение
  3. 3/3

    Спрашивайте

    В чате, или установите CLI и выполните notholy login в терминале.

    $ notholy login