BrassCoders YAML-First Pattern: Feeding Static Analysis Findings to Any AI Coding Assistant

Feeding BrassCoders Output to Any AI Coding Assistant

BrassCoders scans your Python codebase using 12 static-analysis scanners and writes a severity-sorted findings file to .brass/ai_instructions.yaml. This plain YAML file is read by Claude Code automatically, while Cursor and Continue use @file references to include it as context.

Why This Matters

AI coding assistants often lack project-specific awareness of bugs, security issues, or style violations, leading to generic suggestions that ignore existing problems. The ideal model would have persistent, structured context about the codebase’s health without requiring repeated prompting or proprietary plugins. BrassCoders solves this by decoupling the scan from the editor: one offline scan produces a single file any assistant can consume, ensuring findings survive across chat sessions and team members.

Key Insights

• BrassCoders runs 12 static-analysis scanners per scan (2026) — no proprietary encoding, just severity-sorted YAML output.
• YAML-first pattern: scan before opening the editor, then point the assistant at the file so triage starts at confirmation rather than discovery.
• Claude Code reads .brass/ai_instructions.yaml as automatic project context; Cursor requires typing @file .brass/ai_instructions.yaml; Continue uses @file in its chat panel.
• Findings persist between sessions — unlike chat prompts, the file remains until a re-scan overwrites it.

Working Examples

Install BrassCoders and run an offline scan to generate the .brass/ai_instructions.yaml file.

pip install brasscoders
brasscoders --offline scan /path/to/your/project

Practical Applications

• Use case: Teams using Claude Code get automatic project context from .brass/ai_instructions.yaml without extra configuration. Pitfall: Scanning only once — fixed findings remain until you re-run brasscoders —offline scan.
• Use case: Cursor users reference findings via @file .brass/ai_instructions.yaml to address highest-severity items first. Pitfall: The file does not update automatically between scans; stale findings clutter context.
• Use case: Continue teams commit the scanned file to a branch so all members share consistent findings. Pitfall: The @file syntax may change across Continue versions — always check current documentation.

References:

• https://dev.to/coppersundev/feeding-brasscoders-output-to-any-ai-coding-assistant-3j8i