Agents Commenting Rules

Agents Commenting Rules

Apply this skill whenever work touches the repository's agent runtime, tools, or agent workflow wiring.

Scope

Use this skill when editing:

• src/main/agents/**
• adjacent flow files that change agent execution behavior, such as:
• src/main/services/chat-service.ts
• src/main/orpc/router.ts
• src/renderer/src/routes/_dashboard/chat/{-$sessionId}.tsx
• src/renderer/src/lib/orpc.ts

Required Commenting Rule

For touched agent-related code, add or update concise comments on:

• helper functions with non-obvious behavior
• safety guards and fallback paths
• approval, continuation, or retry flow branches
• cross-process or persistence handoff points
• tool execution rules and path-resolution constraints

Keep comments short and specific. Explain the constraint, intent, or reason. Do not narrate obvious assignments or restate the code mechanically.

Good Targets

Typical places that should carry comments after edits:

• path normalization and workspace-boundary checks
• blocked command filters and execution safeguards
• rg to grep fallback logic
• provider/env resolution rules
• assistant message continuation vs new-turn persistence
• approval-requested to approval-responded resume flow

Comment Style

• Prefer one short comment directly above the function or block it explains.
• Focus on "why this branch exists" or "what invariant this preserves".
• Update stale comments in the same change whenever logic moves.
• If a touched helper or flow branch is still hard to understand without reading multiple files, it probably needs a comment.

Before Finishing

For every changed agent-related file:

1. Re-read the touched helper functions and flow branches. 2. Check that non-obvious behavior has a concise comment. 3. Remove or rewrite comments that no longer match the code.