From god-of-debugger
Internal step of /god-of-debugger. Establishes or verifies deterministic repro command before hypothesis generation. Invoked automatically; also directly usable when user says "set up a repro" or "make this reproducible".
How this skill is triggered — by the user, by Claude, or both
Slash command
/god-of-debugger:reproThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Hypothesis meaningless without fireable trigger. This skill outputs one artifact: repro command in session state. Caveman tone. No filler.
Hypothesis meaningless without fireable trigger. This skill outputs one artifact: repro command in session state. Caveman tone. No filler.
Given $ARGUMENTS (bug desc or report path):
pytest -k test_checkout_load, curl -s localhost:3000/api/checkout | jq .status, ./repro.sh.Path: .god-of-debugger/sessions/<session_id>.json
Pointer: .god-of-debugger/current (plain text, active session_id).
session_id = <8-char-uuid>-<branch-slug>. branch-slug = git branch --show-current | tr '/' '-' | tr -cd 'A-Za-z0-9-' truncated 32. Outside git → nobranch.
Minimum fields on first repro:
{
"session_id": "a1b2c3d4-feature-checkout",
"branch": "feature/checkout",
"bug": "<one-sentence restatement>",
"repro": {
"command": "pytest -k test_checkout_load",
"hit_rate": 0.85,
"runs": 20,
"classification": "flaky",
"notes": "fails only after >50 iterations"
},
"hypotheses": [],
"experiments_dir": ".god-of-debugger/experiments/",
"status": "open",
"created_at": "<ISO-8601>",
"closed_at": null
}
.god-of-debugger/current exists + points at open session with repro → print it. Ask replace or keep. Default: keep.session_id, write pointer, create session file.unreliable → status: "repro_unstable". Halt./god-of-debugger:debug."Final line exactly this shape (no extra prose):
Session: a1b2c3d4-feature-checkout
Repro: pytest -k test_checkout_load
Hit: 17/20 (0.85) — flaky
Next: /god-of-debugger:debug
npx claudepluginhub soummyaanon/god-of-debugger --plugin god-of-debuggerCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.