The Agent Isn't the Thing You Design

Описание и применение

В корпоративных AI‑процессах часто ставят агента в центр, что усложняет аудит. Методология ICM заменяет агента структурированными папками и контекстом, позволяя использовать один агент для последовательных этапов с редактируемыми промежуточными файлами.

Задача

Enterprise AI workflows often center on agents, leading to complex, hard‑to‑audit pipelines.

Результат

Using ICM, a single agent can perform multi‑stage workflows with versioned, editable outputs, reducing agent overuse and improving auditability.

Рабочий процесс

• Define a stable reference folder (Layer 3) with rules and constraints.
• Create numbered stage folders (e.g., 01‑research, 02‑draft, 03‑review) each containing a CONTEXT.md contract.
• Run the agent to read the contract and reference folder, generate output into the stage’s output folder.
• Allow a human to edit any output file before the next stage runs.
• Iterate stages sequentially, committing changes to Git for auditability.
• Use a pipeline to automatically run validators between stages.
• Repeat until final output is approved and frozen.

Доказательства

If the prompts and context for each step already live as files in a well‑organized folder tree, one agent reading the right files at the right moment can do the job a multi‑agent framework would do.

Источник

One agent, three stages. Each stage reads its contract plus the shared rules and writes to a folder the next stage reads.

Источник

Every output is an edit surface. A person can open any intermediate file and fix it, and the next stage picks up the fix.

Источник

If the output comes back wrong, fix the contract it read, commit and rerun.

Источник