Stats
Actions
Tags
From payoff
Use when the user asks "did my AI session pay off?", "show my payoff report", "what's my retention rate?", or wants to interpret a payoff HTML report. Runs `payoff report` with the right flags, opens the HTML, and explains the pinpoints + drivers + quadrant in plain English. For fixing wasted time use payoff-waste-triage; for install/config issues use payoff-setup.
How this skill is triggered — by the user, by Claude, or both
Slash command
/payoff:payoff-reportThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Runs and **interprets** the report. Two siblings handle the rest:
Runs and interprets the report. Two siblings handle the rest:
Default: writes self-contained HTML to ~/.claude/payoff/last-report.html
and opens it in the browser.
payoff report --since 7d
Variants:
payoff report --since 30d --by project # monthly, per-project
payoff report --serve --port 7878 # live HTMX-driven server
payoff report --stdout # HTML to stdout (CI / piping)
payoff report --markdown # legacy markdown for terminal
If payoff is not on PATH or hooks aren't installed, hand off to
payoff-setup.
It has four parts, in priority order:
1. "Where time was wasted" (leads the report) — the answer to the most common question. Walk the top 3 pinpoints:
Each pinpoint has a plain-English "Why" column — read it to the user. If they want to act on these rather than just understand them, switch to payoff-waste-triage.
2. Drivers — groups sessions by environment feature (active skills, CLAUDE.md hash, model, edit pattern) with retention/cost deltas vs the all-sessions baseline. Answers "did changing X help?" — always flag it as correlation, not causation.
3. Quadrant — whole-session summary:
| Quadrant | Meaning |
|---|---|
| QUICK WIN | Short session, diff still in HEAD. Cheap value. |
| DEEP VALUE | Long session, diff still in HEAD. Earned its cost. |
| CHEAP WASTE | Short session, diff reverted. Cheap but unproductive. |
| EXPENSIVE WASTE | Long session, diff gone. The signal worth examining. |
Plus three non-scored outcomes: PENDING (window not yet elapsed), REBASED (commit squashed/rebased away), UNMEASURABLE (ran outside git).
4. Totals — token spend, dollar cost, cache hit ratio, sessions by model.
--serve)If the user wants to explore rather than read:
payoff report --serve
HTMX page: click a session row to expand per-file pinpoints, tool-call mix, tokens, cwd; click a driver row to drill into that group's sessions.
Remind the user (the footer says it too):
npx claudepluginhub ayodm/payoff --plugin payoffProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.