bc-fix — Behavioral Contracts Claude Code Skill

WRITES CODE AND CREATES COMMITS. This skill modifies your source files and commits changes to git. It always shows you a full fix plan and waits for your approval before touching anything.

A Claude Code skill that scans your TypeScript codebase for behavioral contract violations, analyzes patterns across all packages, proposes a batched fix plan, and — after your approval — applies fixes with atomic commits.

What it does

1. Runs a full baseline scan (same as bc-scan, read-only)
2. Groups all violations by package and detects cross-cutting patterns
3. Reads every affected file before proposing any changes
4. Presents a complete fix plan and waits for your approval
5. Applies fixes in batches, commits each batch, and rescans after each to verify progress
6. Uploads before/after results to the dashboard so you can track improvement over time

Just want to audit without fixing?

Use bc-scan — the companion skill that is fully read-only. It scans, reports, and uploads results without ever modifying your code or creating commits.

gh repo clone behavioral-contracts/claude-code-skill-bc-scan ~/.claude/skills/bc-scan

Use bc-scan when you want to audit your codebase, run a CI check, or explore violations before deciding what to fix. Use bc-fix when you want to remediate violations and have Claude write and commit the fixes.

Install

gh repo clone behavioral-contracts/claude-code-skill-bc-fix ~/.claude/skills/bc-fix

Or manually:

git clone https://github.com/behavioral-contracts/claude-code-skill-bc-fix ~/.claude/skills/bc-fix

Prerequisites

1. A Behavioral Contracts account — app.behavioral-contracts.com (Solo plan or higher for API access)
2. MCP configured — Follow the setup guide at app.behavioral-contracts.com/developer
3. The repository connected — The GitHub repo you're scanning must be connected in your Behavioral Contracts dashboard

Usage

In any Claude Code session, type:

/bc-fix

Or say: "bc fix", "fix behavioral contract violations", "fix bc errors"

Options

Command	Description
/bc-fix	Scan + fix all ERROR violations (approval required)
/bc-fix --warnings	Also fix WARNING violations
/bc-fix --dry-run	Scan + show plan, but make no changes
/bc-fix --package axios	Fix only violations for one package
/bc-fix --skip-upload	Run locally only, no dashboard upload

Dry run first

If you're unsure what changes will be made, always run --dry-run first:

/bc-fix --dry-run

This shows the full fix plan — which files will be touched, what pattern each fix follows, and how many violations each batch resolves — without writing a single line of code.

How it finds your API key

The skill checks in this order:

1. BC_API_KEY environment variable (if set)
2. ~/.claude.json → projects.<cwd>.mcpServers.behavioral-contracts.headers.Authorization
3. ~/.claude.json → top-level mcpServers.behavioral-contracts.headers.Authorization
4. ~/.claude/claude_desktop_config.json → same path

If MCP is already configured for your project, you're already set.

How fixes are applied

• Fixes match your existing code style (indentation, quote style, import paths)
• Errors are never silently swallowed — fixes always log or rethrow
• Package-specific error type guards are used when available (e.g. axios.isAxiosError())
• Each batch is a separate atomic commit with a descriptive message
• Pre-commit hook failures are handled: the hook error is fixed and a new commit is created (never --amend)

Contributing

Issues and PRs welcome at github.com/behavioral-contracts/claude-code-skill-bc-fix.