Three tools with clear naming (verb_noun pattern) and basic schemas, but significant gaps in parameter descriptions and output documentation. Tool descriptions are adequate (60-90 chars) but lack actionable context for LLM selection. Input schemas use Zod validation but parameter descriptions are minimal or missing. No output schema documentation. Error handling returns isError flag but lacks recovery guidance. The multi-step workflow (blame → location → git context) is well-structured but not reflected in tool composition patterns.

• Document output schemas for all three tools. blame_stack_trace and submit_trace_location return text prompts, declare { type: 'text', content: string }. submit_git_context returns JSON, declare { commitMessage: string, PR: string, comments: string }.
• Expand parameter descriptions with constraints and context. E.g., 'lineRangeStart: Start line number (must be ≥ 1, typically primary_line - 2). Defines the upper bound of the code context window for git blame.'
• Add error recovery guidance. Instead of 'Invalid location: ...', return 'Invalid location: filePath must exist in the repo. Verify the path is relative to the repo root (e.g., src/index.ts, not /absolute/path/src/index.ts).'
• Refactor the workflow to encapsulate git operations inside the tools, not in LLM prompts. submit_trace_location should accept filePath + lineRange and internally call git blame, returning structured results. This eliminates the manual git command step.
• Add tool annotations (readOnlyHint) to all three tools since they only read stack traces and git history, no state modification. This signals to agents that these calls are safe to retry.