Skip to content

[feature] Agent Proxy: Lint Shielding and Auto-Remediation Sub-Agent #60

Description

@paudley

Pre-flight

  • I searched existing issues first.

Problem statement

Expensive main coding agents waste massive amounts of tokens and API time in a loop trying to fix trivial syntax errors, indentation mistakes, trailing whitespaces, or missing imports after generating code.

Proposed solution

When the agent pushes a file edit, the proxy should temporarily intercept the write and run coding-ethos-lint. If simple, resolvable formatting or syntax errors are detected, the proxy should silently fix them and apply the change. The proxy only returns an error back to the main agent if the logic is fundamentally broken, shielding the expensive LLM from trivial syntax loops.

Examples or prior art

Advanced agent architectures use a cheap/fast model or standard static analysis as a "sub-agent" to fix syntax before the main agent even knows it made a mistake.

Alternatives considered

Piping raw linter output back to the LLM, which relies on the LLM to perfectly execute the fix in its next turn (token-heavy).

Additional context

This relies heavily on the existing auto-fix and rule logic within coding-ethos-lint and requires the proxy to reliably manage file states.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions