Correctless

/cmetrics — Workflow Metrics Dashboard

Aggregate all workflow data into a project health and ROI dashboard showing bugs caught, token cost, and trends.

When to Use

Monthly, to track how the workflow is performing over time.
Before presenting to stakeholders, to justify the workflow investment with data.
When deciding whether to increase workflow intensity.
When something feels off — metrics reveal which phases are pulling weight and which are not.
Not for: Single-feature summaries (use /csummary), or checking current workflow state (use /cstatus).

How It Fits in the Workflow

This skill is standalone and project-wide. It reads accumulated data from every feature that has gone through the Correctless workflow — QA findings, verification reports, antipatterns, drift debt, audit history, token logs, and git history. It produces a comprehensive dashboard with actionable health analysis. Run it periodically to spot trends.

What It Does

Counts features completed, total issues caught (deduplicated across phases), and bug escape rate.
Breaks down issues by phase (Review, Test Audit, QA, Verify, Audit) to show where value concentrates.
Tracks antipattern growth by category, flagging systemic issues that may need architectural fixes.
Monitors drift debt health: open items, staleness, accumulation vs. resolution rate.
Token ROI Analysis: Reads .correctless/artifacts/token-log-*.json to compute cost per bug caught, tokens per feature by phase, tokens per LOC, and estimated production fix cost avoided. Shows whether token spend is efficient (e.g., “65% of tokens go to TDD, which catches 60% of bugs”).
Session Analytics: Reads Claude Code session-meta and facets data from ~/.claude/usage-data/, filtered to the current project. Reports exact token cost (ground truth), tool distribution, friction rate, user engagement, and outcome rate.
Correctless vs Freeform comparison: Identifies which sessions used Correctless (by checking artifact timestamps and tool patterns) vs. freeform coding, then compares outcome rate, friction, duration, and token usage between the two groups.
Cross-metric correlation analysis: flags patterns like “specs revised mid-TDD frequently AND antipattern growth accelerating in the same category.”

Example

User: /cmetrics

# Correctless Metrics — my-project
# Generated: 2026-03-29

## Overview
- Features completed: 8
- Total issues caught: 47
- Bug escape rate: 2 escaped / 47 caught (4.1%)
- Workflow active since: 2026-01-15

## Issues by Phase
| Phase        | Issues | % of Total | Notes                              |
|--------------|--------|------------|------------------------------------|
| Review       | 12     | 25%        | Security checklist added 40% of these |
| QA           | 22     | 47%        | 5 class fixes added structural tests  |
| Verify       | 8      | 17%        |                                    |
| Audit        | 5      | 11%        |                                    |

## Token ROI Analysis
- Total tokens tracked: 2.4M across 8 features
- Cost per bug caught: ~51k tokens (47 bugs)
- Estimated production fix cost avoided: 47 bugs x 2-10 hrs = 94-470 hrs

## Session Analytics
- Correctless sessions: 62% fully achieved outcome
- Freeform sessions: 41% fully achieved outcome

## Health Analysis
- QA rounds trending down (3.1 -> 2.3 avg) — workflow getting more effective.
- Error handling antipatterns growing fastest (+4 in 3 months). Consider architectural pattern.
- Spec revision rate: 25% of features — acceptable.

What It Reads / Writes

Reads	Writes
`.correctless/artifacts/qa-findings-*.json`	`.correctless/artifacts/metrics-{date}.md`
`docs/verification/*-verification.md`
`.correctless/meta/workflow-effectiveness.json`
`.correctless/antipatterns.md`
`.correctless/meta/drift-debt.json`
`.correctless/artifacts/findings/audit-*-history.md`
`.correctless/specs/*.md`
`.correctless/artifacts/summary-*.md`
`.correctless/artifacts/token-log-*.json`
`.correctless/artifacts/audit-trail-*.jsonl`
`~/.claude/usage-data/session-meta/*.json`
`~/.claude/usage-data/facets/*.json`
`docs/decisions/*.md`
Git log

Intensity Levels

Standard intensity: Omits Olympics convergence analysis and Olympics history table (those features require high+ intensity).
High/critical intensity: Adds Olympics history, convergence speed analysis, and audit-specific metrics.

Common Issues

Sparse data on new projects: This is expected. Data accumulates with each feature. The dashboard reports what exists and notes what will appear after more features run.
Session-meta not found: Claude Code session data may not exist for the project yet. Session analytics will appear after a few sessions.
Token logs vs. session-meta divergence: If the numbers differ significantly, the dashboard notes it: the gap is orchestrator overhead not captured by subagent tracking.

This site is open source. Improve this page.