feat: Add get_error_details() for tool failure analysis

[evansenter]

Summary

• Add get_error_details() MCP tool to show which specific parameters caused errors
• Add error-details CLI command
• Update documentation

Problem

analyze_failures() shows "932 Glob errors" but not which patterns failed. Without this, workflow improvements can only identify that errors happen, not why.

Solution

Use SQLite's json_extract() on existing tool_input_json column - no schema changes needed.

Example output:

Glob errors by pattern:
  "*" - 922 errors (rust-genai)  ← problematic pattern found!
  "**/*.rs" - 6 errors

Test plan

• All 293 tests pass
• CLI error-details command works
• MCP tool functional

Closes #60

🤖 Generated with Claude Code

[claude]

Prompt: evansenter/dotfiles/.../claude-review.md

Code Review

Summary

This PR adds a new get_error_details() MCP tool and error-details CLI command to show which specific parameters (patterns, commands, files) caused tool failures, enabling drill-down from aggregate error counts to actionable specifics.

Issues Found

Critical

None

Important

• Missing tests for query_error_details(): The new query function in queries.py:1904-2007 has no corresponding tests. Other query functions like query_tool_frequency have tests in test_queries.py. This is a new public function that should be tested to verify:
  • Basic error aggregation works
  • Tool filter parameter works
  • Limit parameter works correctly
  • The join between tool_use and tool_result events is correct

Suggestions

• src/session_analytics/cli.py:807-817 - No custom CLI formatter registered for error-details output. Without a formatter, the command will output raw JSON even without --json. Consider adding a @_register_formatter similar to _format_failures (line 381) for better human-readable output.
• Partial implementation of RFC requirements: The linked issue RFC: Add get_error_details() for tool failure analysis #60 proposed group_by parameter and sample_error field in output, which are not implemented. While the current implementation is useful, documenting that these were intentionally deferred would help future maintainers.

Verdict

REQUEST_CHANGES - Missing tests for the new query function is an important gap that should be addressed before merge.

---

Automated review by Claude Code

[claude]

Prompt: evansenter/dotfiles/.../claude-review.md

Code Review

Summary

This PR adds a get_error_details() MCP tool and error-details CLI command that enables drill-down from aggregate error counts (e.g., "932 Glob errors") to actionable specifics (e.g., "pattern '*' failed 922 times in rust-genai"). The implementation uses SQLite's json_extract() on existing tool_input_json columns with no schema changes required.

Previous Review Status

The previous review's Important issue (missing tests) has been fully addressed - comprehensive tests were added in TestQueryErrorDetails class covering: basic error aggregation, tool filtering, limit parameter, file path errors, Grep with search_path, empty results, and days filtering.

Issues Found

Critical

None

Important

None

Suggestions

• src/session_analytics/cli.py:807-817 - No custom CLI formatter registered for error-details output. Without a formatter, the command outputs raw JSON (nested dict structure) even without --json flag. Other similar commands like failures have formatters (line 381) for human-readable output. Consider adding a @_register_formatter that formats errors grouped by tool, e.g.:

  Error Details (last 7 days)
  Total errors: 928
  
  Glob (922 errors):
    "*": 917 errors (rust-genai)
    "**/*.rs": 5 errors (rust-genai)
  
  Bash (6 errors):
    "git status": 6 errors (dotfiles)
  

Verdict

REQUEST_CHANGES - One suggestion for improved CLI UX.

---

Automated review by Claude Code

[claude]

Prompt: evansenter/dotfiles/.../claude-review.md

Code Review

Summary

This PR adds a get_error_details() MCP tool and error-details CLI command that enables drill-down from aggregate error counts to actionable specifics (e.g., which Glob pattern failed 922 times). Uses SQLite's json_extract() on existing tool_input_json column with no schema changes required.

Previous Review Status

Both issues from prior reviews have been addressed:

• Missing tests: Comprehensive test suite added in TestQueryErrorDetails class (7 test methods covering aggregation, filtering, limits, file paths, search paths, empty results, and date filtering)
• Missing CLI formatter: _format_error_details() formatter is now present at cli.py:404-437

Issues Found

Critical

None

Important

None

Suggestions

None

Verdict

APPROVE - Code looks good, no issues found. The implementation follows project conventions (CLI/MCP parity, proper storage API usage), includes thorough tests, and the documentation updates in guide.md and README.md are accurate.

---

Automated review by Claude Code