Reduce context noise
Exclude generated artifacts, dependency folders, logs, and legacy areas that distract retrieval and reasoning.
Draft v0.1 standard for AI coding tools
One file. One syntax. Every agent.
AgentIgnore defines a vendor-neutral .agentignore file for excluding project files from passive AI coding agent discovery, indexing, retrieval, context construction, and automated modification.
AgentIgnore gives projects a single place to express which files should stay out of passive agent context.
Exclude generated artifacts, dependency folders, logs, and legacy areas that distract retrieval and reasoning.
Reduce accidental inclusion of irrelevant files in passive discovery while preserving explicit user-directed access semantics.
Define behavior that can work across all major agentic coding tools and future ecosystems.
AgentIgnore is intended to reduce accidental exposure to AI coding agents during passive operations. Sensitive files still require proper access control, encryption, secret management, and repository hygiene.
AgentIgnore does not claim invention of the idea. The project exists to help existing efforts and tool-specific requests converge on a single vendor-neutral standard.
And many more across agentic coding tools and adjacent ecosystems.
The current proposal is intentionally narrow: recognize a root-level .agentignore, apply familiar ignore patterns, and keep ignored paths out of passive agent operations.
.agentignoreList files and directories that agents should not passively discover, index, retrieve, summarize, or modify.Use the RFC for the current proposal, examples, rationale, and open design space.
AgentIgnore is a draft proposal. Review the RFC, report compatibility findings, or suggest focused changes through GitHub.