Two ready-made agent skills that drive the oopsie CLI for you. Works with Claude Code, Codex, Hermes, OpenClaw, and any other AI coding agent that reads skills from ~/.claude/skills/ — or just point your agent at the markdown files directly.

Install both

One copy-paste — installs oopsie and oopsies-triage-and-fix for any agent that reads ~/.claude/skills/:

mkdir -p ~/.claude/skills/oopsie ~/.claude/skills/oopsies-triage-and-fix
curl -fsSL https://raw.githubusercontent.com/theinventor/Oopsie/main/skills/oopsie/SKILL.md \
  -o ~/.claude/skills/oopsie/SKILL.md
curl -fsSL https://raw.githubusercontent.com/theinventor/Oopsie/main/skills/oopsies-triage-and-fix/SKILL.md \
  -o ~/.claude/skills/oopsies-triage-and-fix/SKILL.md

Skills require the oopsie CLI to be installed and at least one connection configured (oopsie config add ...).

/oopsie

The everyday driver. Ask Claude to "check oopsie", "show what's broken", or "show me error #42" and this skill runs the right CLI command, reads the stack trace, jumps to the offending file, and helps you fix it.

$ claude
> check oopsie

[runs oopsie whoami, oopsie errors --status unresolved]
3 unresolved errors in myapp:
  #47  NoMethodError — 1.2k occurrences, 2m ago
  #45  ActiveRecord::RecordNotFound — 89 occurrences
  #42  ArgumentError — 12 occurrences

View source →

/oopsies-triage-and-fix

The end-to-end button. Matches the current repo to an Oopsie project, picks the most urgent error (or asks if it's a coin flip), claims it with workflow state so nobody else duplicates the work, leaves evidence notes, fixes it, then runs /review, /qa, /ship, and /land without stopping.

$ claude
> triage oopsies

[matches repo to project, picks #47]
Setting #47 in_progress and starting the fix...
[fix → review → qa → ship → land]

View source →

How to use

Most agents pick up skills from ~/.claude/skills/ on launch. After installing, just talk to your agent normally:

If your agent uses a different skills directory, adjust the install paths to match. The skills are plain markdown — any agent that can follow instructions can use them.

No CLI? Grab it here first — the skills are thin wrappers around it.