Dogfooding
SLOPE scores its own development.
Every sprint since S1 has a scorecard — 69 sprints of real work, all public. The metrics on this site are computed from real scorecards, not synthetic benchmarks.
View every scorecard on GitHub →What SLOPE Is
Retrospectives with data instead of feelings.
Every sprint gets a scorecard. Over time, scorecards reveal patterns — which types of tasks get over-scoped, where hazards cluster, how estimation accuracy trends. SLOPE is the structured feedback loop that turns ad-hoc AI coding sessions into a measurable, improving process.
How It Works
Six layers of structured feedback.
Scorecards
Each sprint records approach complexity, outcomes, and hazards per ticket. The raw data behind every insight.
Handicap tracking
Rolling performance metrics reveal trends across sprints — estimation accuracy, delivery rate, miss patterns.
Guard hooks
29 real-time checks that inject context, warnings, and blocks into your agent's workflow as you code.
Workflow engine
Structured sprint execution with plan review gates, completion conditions, and pause/resume. sprint-standard is the default path for every sprint.
Multi-repo & teams
slope org aggregates metrics across repos. slope session coordinates multiple agents with handoff protocols and live dashboards.
Pluggable metaphors
Same scoring engine, 7 vocabulary options — golf, agile, tennis, baseball, gaming, D&D, matrix.
Who It's For
Built for people shipping with AI agents.
Developers
Using AI coding agents — Claude Code, Cursor, Windsurf, Codex CLI, Pi.dev, Cline, OB1, OpenCode — who want measurable improvement over time.
Tech leads and PMs
Who need visibility into AI-assisted development quality without reading every diff.
Teams
Building processes around AI pair programming who want consistency across agents and sessions.
The Numbers
Live stats from the reference implementation.
Sprints
Handicap
Tests
Guards
CLI Commands
Metaphors