diagnose

Name: diagnose
Author: eduwxyz

15stars1forks

How to install diagnose?

eduwxyz/my-awesome-skills/diagnose

$npx -y skills add eduwxyz/my-awesome-skills --skill diagnose

Installs into the current project.

Summary

Turn a bug symptom into a fix with a regression test that locks it down. If the user opens vague ("there's a bug", "/diagnose"), interview them one question at a time until you have enough to attempt a reproduction; if you still cannot reproduce, say so explicitly. Then drive the

Files · 1

View on GitHub

SKILL.md

1	# Diagnose
2
3	Symptom in, fix-with-regression-test out. An interview if the user opens vague, then five phases, two cardinal rules.
4
5	## Skip diagnose for
6
7	One-line obvious fixes (typo, missing import, off-by-one in a test you just wrote). "Bugs" that are actually missing functionality — that's a feature, use `interview-to-spec`. Flaky tests caused by infrastructure (CI, network) rather than code under test.
8
9	## Two rules that keep diagnosis honest
10
11	1. No loop, no Phase 2+. If you cannot reproduce the bug deterministically, you cannot hypothesise — you'll just guess.
12	2. One variable at a time during probing. Changing two things at once tells you nothing about either.
13
14	If either slips, back up.
15
16	## Phase 0 — Understand the bug
17
18	Before building a loop, you need to know what you're trying to reproduce. If the user opened with detail covering the points below, skip this phase. If they opened vague ("there's a bug", "X is broken", or just `/diagnose`), interview them.
19
20	Ask one question at a time. Stop the moment you have enough to attempt Phase 1 — do not gather more than you need.
21
22	The minimum set you're trying to fill:
23
24	- Expected vs. actual. What should have happened? What actually happened?
25	- Trigger. What were they doing when it happened? Can they make it happen again on demand, or is it intermittent?
26	- Recency. Did this used to work? When did it start failing? What changed recently (deploy, dependency, data, config)?
27	- Environment. Where? (prod / staging / local; specific browser, OS, account, dataset.)
28	- Evidence. Any error message, stack trace, log line, screenshot, or HAR they can paste?
29
30	If the answer to a question is in the codebase rather than in the user's head, read the code instead of asking.
31
32	When the answers converge, propose a short slug for the bug (e.g. `token-expiry-not-rejected`). This is the filename for the spec written in Phase 5. Then move to Phase 1.
33
34	If after the interview the user still cannot give you enough to reproduce, that is itself a Phase 1 outcome — go to Phase 1 and follow When you genuinely cannot build a loop.
35
36	## Phase 1 — Build the feedback loop
37
38	This is the skill. Everything downstream is mechanical once you have a fast, deterministic, agent-runnable pass/fail signal for the bug. Without one, no amount of staring at code finds the cause.
39
40	Spend disproportionate effort here. Be aggressive. Refuse to give up.
41
42	### Ways to build one — try in roughly this order
43
44	1. Failing test at whatever seam reaches the bug — unit, integration, e2e.
45	2. Curl / HTTP script against a running dev server.
46	3. CLI invocation with a fixture input, diffing stdout against a known-good snapshot.
47	4. Headless browser (Playwright / Puppeteer) — drives the UI, asserts on DOM/console/network.
48	5. Replay a captured trace. Save a real request/payload/event log to disk; replay it through the code path in isolation.
49	6. Throwaway harness. A minimal subset of the system that exercises the bug code path with one function call.
50	7. Property / fuzz loop. For "sometimes wrong output", run 1000 random inputs and look for the failure mode.
51	8. Bisection harness. If the bug appeared between two known states (commit, dataset, version), automate "boot at state X, check, repeat" so `git bisect run` works.
52	9. Differential loop. Run the same input through old vs new (or two configs) and diff outputs.
53	10. HITL bash script. Last resort. If a human must click, drive them with a structured loop and capture output back to you.
54
55	### Iterate on the loop itself
56
57	Treat the loop as a product. Once you have a loop, ask:
58
59	- Faster? Cache setup, skip unrelated init, narrow scope.
60	- Sharper signal? Assert on the specific symptom, not "didn't crash".
61	- More deterministic? Pin time, seed RNG, isolate filesystem, freeze network.
62
63	A 30-second flaky loop is barely a loop. A 2-second deterministic loop is a debugging superpower.
64
65	### Non-deterministic bugs
66
67	Goal isn't a clean repro — it's a higher reproduction rate. Loop the trigger 100×, parallelise, narrow timing windows, inject sleeps. A 50%-flake bug is debuggable; 1% is not — keep raising the rate until it is.
68
69	### When you genuinely cannot build a loop
70
71	Stop and say so explicitly. List what you tried. Ask the user for one of:
72
73	- Access to an envi

People who liked this also viewed

6 picks

Type

diagnose

15stars1forks

View on GitHub

How to install diagnose?

eduwxyz/my-awesome-skills/diagnose

$npx -y skills add eduwxyz/my-awesome-skills --skill diagnose

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Diagnose
2
3	Symptom in, fix-with-regression-test out. An interview if the user opens vague, then five phases, two cardinal rules.
4
5	## Skip diagnose for
6
7	One-line obvious fixes (typo, missing import, off-by-one in a test you just wrote). "Bugs" that are actually missing functionality — that's a feature, use `interview-to-spec`. Flaky tests caused by infrastructure (CI, network) rather than code under test.
8
9	## Two rules that keep diagnosis honest
10
11	1. No loop, no Phase 2+. If you cannot reproduce the bug deterministically, you cannot hypothesise — you'll just guess.
12	2. One variable at a time during probing. Changing two things at once tells you nothing about either.
13
14	If either slips, back up.
15
16	## Phase 0 — Understand the bug
17
18	Before building a loop, you need to know what you're trying to reproduce. If the user opened with detail covering the points below, skip this phase. If they opened vague ("there's a bug", "X is broken", or just `/diagnose`), interview them.
19
20	Ask one question at a time. Stop the moment you have enough to attempt Phase 1 — do not gather more than you need.
21
22	The minimum set you're trying to fill:
23
24	- Expected vs. actual. What should have happened? What actually happened?
25	- Trigger. What were they doing when it happened? Can they make it happen again on demand, or is it intermittent?
26	- Recency. Did this used to work? When did it start failing? What changed recently (deploy, dependency, data, config)?
27	- Environment. Where? (prod / staging / local; specific browser, OS, account, dataset.)
28	- Evidence. Any error message, stack trace, log line, screenshot, or HAR they can paste?
29
30	If the answer to a question is in the codebase rather than in the user's head, read the code instead of asking.
31
32	When the answers converge, propose a short slug for the bug (e.g. `token-expiry-not-rejected`). This is the filename for the spec written in Phase 5. Then move to Phase 1.
33
34	If after the interview the user still cannot give you enough to reproduce, that is itself a Phase 1 outcome — go to Phase 1 and follow When you genuinely cannot build a loop.
35
36	## Phase 1 — Build the feedback loop
37
38	This is the skill. Everything downstream is mechanical once you have a fast, deterministic, agent-runnable pass/fail signal for the bug. Without one, no amount of staring at code finds the cause.
39
40	Spend disproportionate effort here. Be aggressive. Refuse to give up.
41
42	### Ways to build one — try in roughly this order
43
44	1. Failing test at whatever seam reaches the bug — unit, integration, e2e.
45	2. Curl / HTTP script against a running dev server.
46	3. CLI invocation with a fixture input, diffing stdout against a known-good snapshot.
47	4. Headless browser (Playwright / Puppeteer) — drives the UI, asserts on DOM/console/network.
48	5. Replay a captured trace. Save a real request/payload/event log to disk; replay it through the code path in isolation.
49	6. Throwaway harness. A minimal subset of the system that exercises the bug code path with one function call.
50	7. Property / fuzz loop. For "sometimes wrong output", run 1000 random inputs and look for the failure mode.
51	8. Bisection harness. If the bug appeared between two known states (commit, dataset, version), automate "boot at state X, check, repeat" so `git bisect run` works.
52	9. Differential loop. Run the same input through old vs new (or two configs) and diff outputs.
53	10. HITL bash script. Last resort. If a human must click, drive them with a structured loop and capture output back to you.
54
55	### Iterate on the loop itself
56
57	Treat the loop as a product. Once you have a loop, ask:
58
59	- Faster? Cache setup, skip unrelated init, narrow scope.
60	- Sharper signal? Assert on the specific symptom, not "didn't crash".
61	- More deterministic? Pin time, seed RNG, isolate filesystem, freeze network.
62
63	A 30-second flaky loop is barely a loop. A 2-second deterministic loop is a debugging superpower.
64
65	### Non-deterministic bugs
66
67	Goal isn't a clean repro — it's a higher reproduction rate. Loop the trigger 100×, parallelise, narrow timing windows, inject sleeps. A 50%-flake bug is debuggable; 1% is not — keep raising the rate until it is.
68
69	### When you genuinely cannot build a loop
70
71	Stop and say so explicitly. List what you tried. Ask the user for one of:
72
73	- Access to an envi

People who liked this also viewed

6 picks

Type