cocowatch

Observes developer engagement patterns and classifies collaboration zones at session checkpoints.

720|83|Updated Dec 8, 2024

Search Tags:#checkpoint #session analytics #engagement tracking #Code review #Observability #Developer workflow #cocoplus

Ready to run

npx skills add https://github.com/Snowflake-Labs/cocoplus --skill cocowatch

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: cocowatch
Source: https://github.com/Snowflake-Labs/cocoplus/tree/main/.cortex/skills/cocowatch
Command: npx skills add https://github.com/Snowflake-Labs/cocoplus --skill cocowatch

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Teams using AI-assisted development pipelines lack visibility into how actively developers review and engage with generated outputs, making it hard to detect over-delegation or under-utilization of the system.

Core Features & Use Cases

Delegation Intensity Tracking: Records the ratio of outputs accepted without modification versus modified before acceptance across checkpoint reviews.
Review Depth Measurement: Captures acknowledgment timing, reasoning text, and handling of BLOCKING findings as raw engagement signals.
Engagement Zone Classification: Synthesizes observations into Zone 1 (over-delegation), Zone 2 (healthy collaboration), or Zone 3 (under-utilization) summaries surfaced only at $ship and FULL checkpoints.
Use Case: During a long CocoPlus build session, a lead wants to know whether reviewers are rubber-stamping SecondEye findings; CocoWatch logs each acknowledgment and surfaces an advisory profile at ship time.

Quick Start

Ask the agent to show the CocoWatch session collaboration profile at the next FULL checkpoint.

Frequently Asked Questions about cocowatch

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I monitor developer engagement in AI-assisted coding sessions?▼

CocoWatch runs as a non-blocking observer that logs each checkpoint approval, finding acknowledgment, and classification override with timestamps and modification flags. It surfaces a collaboration profile summary only at $ship and FULL checkpoints.

What is an engagement zone in session observability?▼

An engagement zone classifies collaboration patterns into three types: Zone 1 over-delegation, Zone 2 healthy collaboration, and Zone 3 under-utilization. The classification is derived from delegation intensity and review depth signals and is advisory only.

Can CocoWatch block or gate a pipeline based on low engagement?▼

No. CocoWatch is structurally non-blocking via its frontmatter and must never influence any gate, classification, or recommendation. It is purely observational and surfaces summaries only at designated checkpoints.

Where are CocoWatch session observations stored?▼

Observations are written to .cocoplus/lifecycle/cocowatch-session.md as a running log. The file is session-ephemeral, excluded from git via .cocoplus/.gitignore, and overwritten at each new session start.

What are the limitations of engagement observation tools?▼

Engagement signals are context-dependent; short sessions naturally show lower modification rates, so uniform thresholds do not apply. Rapid approvals are not proof of poor engagement without session context, and classifications remain advisory rather than prescriptive.