technical-premortem

technical-premortem is a command for coding agents from nicelight/DevRails-26. It costs 60 tokens per session (1,890 once invoked), scanned A, original, MIT.

A read-only technical pre-mortem: an early review that assumes a planned change has failed, then works backward to find likely causes and affected parts.

In plain words
What is it for?
Use it to inspect a task or implementation plan, map its affected callers, data, integrations, and operational owners, and define conditions for proceeding safely.
Why use it?
It exposes risks and unclear safety conditions before implementation, without changing requirements, approvals, tasks, or project files.

Command

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 commands/nicelight/devrails-26/technical-premortem
Clone the repo
git clone --depth 1 https://github.com/nicelight/DevRails-26

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 technical-premortem

README.md
[![agentmods](https://agentmods.dev/badge/commands/nicelight/devrails-26/technical-premortem.svg)](https://agentmods.dev/commands/nicelight/devrails-26/technical-premortem)
Your own site
<a href="https://agentmods.dev/commands/nicelight/devrails-26/technical-premortem"><img src="https://agentmods.dev/badge/commands/nicelight/devrails-26/technical-premortem.svg" alt="Measured on agentmods" height="20"></a>
Per session 60 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,890 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 $0.00060 $0.01890
Opus 5 $0.00030 $0.00945
Sonnet 5 $0.00012 $0.00378
Haiku 4.5 $0.00006 $0.00189

Measured 4d ago against content hash 1810ccbafb63, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

technical-premortem 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 4d 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/_shared/references/commands/technical-premortem.md · 170 lines

How it starts

The opening of the file, as written. The whole thing — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/technical-premortem — Technical pre-mortem

Skill advisory и read-only. Он не является workflow gate, не заменяет /review-feat-plan, /review-tasks-plan, /verify, /red-verify или /mb-doctor и не меняет task lifecycle, approvals, ownership или durable artifacts.

Finding не создаёт requirement и не расширяет accepted target. Если нужная mitigation меняет product behavior, design, task boundary, tier или proof contract, направь её существующему owning workflow.

<input_contract> Expected $ARGUMENTS: один TASK-NNN-TN-FT-NNN-WN либо однозначно названный implementation plan или planned change. Без аргумента используй текущий operator context только когда target однозначен.

Прочитай относящиеся к target task/plan, accepted specs, code, tests, configuration, migrations и доступный operational context. Восстанови:

  • что меняется: files, modules, behavior, data/schema/config и contracts;
  • кто зависит: callers, consumers, jobs, integrations и operational owners;
  • что разделяется: state, cache, sessions, queues, storage и environment;
  • intended outcome, compatibility constraints, proof path и accepted limits.

Если material scope, intended behavior или authoritative target не восстановимы, задай короткие точечные вопросы и остановись без verdict. Не выдумывай структуру, потребителей, production conditions или требования. </input_contract>

<failure_frame> Представь, что изменение уже смёржено и, где применимо, задеплоено. Проверь каждый относящийся к target путь провала:

  1. Само не сработало — intended behavior не достигнут из-за логики, данных, edge case или неверного допущения.
  2. Сломало соседнее — регрессировал consumer, contract, shared state или скрытая зависимость.
  3. Ухудшило, не сломавшись явно — деградировали security, performance, reliability, operability, maintainability или architecture boundary.
  4. Сработало, но навредило позже — harm проявился на реальных данных, под нагрузкой, при конкуренции, после deploy или rollback.

Не останавливайся на первом очевидном сценарии. Сначала найди конкретные candidate failures по применимым путям и surfaces, затем отфильтруй их по evidence, consequence и recoverability. Отклонённые speculative candidates в ответ не включай. </failure_frame>

<failure_surfaces> Используй только относящиеся к target surfaces, но не пропускай применимые:

  • correctness и historical/null/partial data;
  • callers, contracts, integrations, versions и shared state;
  • migrations, irreversible writes, queues, caches и rollback residue;
  • security, permissions, secrets и trust boundaries;
  • dependency failure, retries, timeouts, idempotency и concurrency;
  • production scale, latency, memory, deployment order и config drift;
  • observability: как оператор или тест узнает о конкретном сбое;
  • architecture/maintainability: какой accepted owner, boundary или invariant будет нарушен;
  • agent-execution error: где исполнитель может двусмысленно прочитать target, выдумать API, затронуть лишнее, нарушить accepted boundary или предъявить нерелевантное evidence.

Agent-execution surface проверяется всегда; finding публикуется только при конкретном material механизме или decision-relevant blind spot. </failure_surfaces>

<risk_model> Классифицируй только findings, способные изменить решение или pre-flight:

  • 🐅 Tiger — подтверждённый material risk: evidence связывает конкретный mechanism с наблюдаемым symptom и affected surface. Игнорирование способно сорвать accepted outcome. Укажи urgency merge_blocker, follow_up или observe.
  • 🐯 Paper Tiger — сценарий выглядит опасно, но inspected evidence, ограниченный impact или дешёвая recoverability показывают, почему он не blocker. Включай его только когда явное понижение риска предотвращает ненужную сложность или влияет на verdict.
  • 🐘 Elephant — существенное допущение, blind spot или отсутствие evidence, на котором зависит решение. Назови, какое решение от него зависит, какой evidence отсутствует и кто может закрыть неизвестность.

Read the full file on GitHub · 170 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. 4d ago First seen · 170 lines · 60 tokens per session scan A 1810ccbafb63

Subscribe to this mod's changes

technical-premortem is a command published in the GitHub repository nicelight/DevRails-26 (22 stars, last pushed 6d ago), licensed MIT. It adds 60 tokens to every session and 1,890 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.