โ† All blog posts

AI agents

Custom agent instructions

Replace the built-in agent workflow with your own preamble in Settings, your rules govern every session.

Feature 39 of 72FeatureAI agents, built in3 min read
Abstract ForkMesh artwork for Custom agent instructions

Why it matters

Custom agent instructions gives ForkMesh a concrete mechanism for project-specific agent rules controlled from settings. Replace the built-in agent workflow with your own preamble in Settings, your rules govern every session.

AI coding only helps when the agent can work in a real repository without trampling human state or hiding what it did. In ForkMesh, custom agent instructions addresses that need at the feature level, keeping the behavior close to the repository instead of buried in a detached hosted layer.

How ForkMesh handles it

ForkMesh handles custom agent instructions through the same split it uses across the product: ForkMesh runs agents beside the repository, capturing transcripts, branches, diffs, prompts, usage, questions, and handoffs in the desktop app.

The behavior is intentionally narrow: Replace the built-in agent workflow with your own preamble in Settings, your rules govern every session. Humans keep control of instructions, answers, review, and merge decisions.

Title focus: Custom agent instructions stays tied to project-specific agent rules controlled from settings and the local-first repository model.

Where it fits

Use custom agent instructions when agent work should be reviewable, resumable, and connected to the same repo UI humans use.

It pairs with the rest of AI agents, built in because Custom agent instructions keeps project-specific agent rules controlled from settings connected to the repository, its signatures, and the nodes that serve it.