agentleFS
UwayxtUwayxt/agent-skills/.agents/skills/ux-chaos-monkey/SKILL.md

ux-chaos-monkey

Skill4 starsChanged 3 months agoMITView on GitHub

Inject destructive chaos vectors — extreme text, edge-case data, and flaky network conditions — to stress-test UI resilience before real users find the breaks. Use when stress test, chaos test, text overflow, data edge case, network failure, layout break, or "what happens when…" questions arise.

  • Deletes or force-pushes

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Not re-testing after auto-fixes (fix may introduce new overflow elsewhere)

    line 245

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

🎭 UX Chaos Monkey & Edge-Case Stress Testing

Thesis

A UI that only works with clean, predictable data is not production-ready. Real users will paste Russian Wikipedia articles into search fields, have 120-character names, and lose their internet connection at the worst possible moment.

The Chaos Monkey doesn't look for bugs — it engineers controlled catastrophes to prove the interface survives them. If it breaks in this controlled test, it will break in production.


When to Activate

Activate automatically when the user mentions:

  • "stress test", "chaos test", "what happens when the data is weird"
  • "text overflow", "layout breaks", "card breaks with long text"
  • "extreme values", "edge case", "what if the name is very long"
  • "network failure", "offline during", "what if it disconnects"
  • "data injection", "fuzz test", "boundary test"

3-Protocol Execution System

Execute all 3 protocols in sequence. Never skip a protocol. Report all findings at the end as a unified Chaos Report.


─── PROTOCOL 1: TEXT EXPLOSION TEST

Objective: Prove that no text content — regardless of length or language — causes horizontal overflow, layout rupture, or obscured interactive elements.

1.1 — Identify Injection Targets

Scan the UI and flag every component that renders user-supplied or dynamic text:

Target ComponentRisk LevelCritical Check
Button / CTA label🔴 HIGHDoes the button grow wider than its container?
Table cell / Data grid🔴 HIGHDoes the row height explode or overflow horizontally?
Badge / Tag / Chip🔴 HIGHDoes the badge break out of its pill container?
Card title / Heading🟡 MEDDoes the card resize correctly or clip at 2 lines?
Breadcrumb item🟡 MEDDoes the breadcrumb trail collapse gracefully?
Tooltip content🟡 MEDDoes the tooltip overflow the viewport edge?
Input placeholder🟢 LOWDoes the placeholder ellipsize before the field?
Notification message🟢 LOWDoes the toast/snackbar clip at max-width?

1.2 — Text Bomb Library

Inject each of the following strings into every HIGH-risk target:

BOMB-A: German Compound Noun (45 chars)

Rechnungsabgrenzungsposten
Kraftfahrzeughaftpflichtversicherung
Donaudampfschifffahrtsgesellschaftskapitän

BOMB-B: Russian Inflected Word (20–24 chars)

достопримечательности
железнодорожный
непосредственно

BOMB-C: Finnish Agglutinative (58 chars)

lentokonesuihkuturbiinimoottoriapumekaanikkoaliupseeri

BOMB-D: Zero-Width / No-Break String (40 repeated chars — no natural word-break)

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
อาหารไทยอร่อยมากๆเลยนะครับ

BOMB-E: 300% Oversize Label

Submit Order Confirmation and Send Invoice to Primary Billing Contact and CC Secondary Approver

1.3 — Pass / Fail Criteria

For each injection, verify ALL of the following:

  • Zero horizontal scroll introduced at any viewport width ≥ 320px
  • text-overflow: ellipsis fires before overflow occurs
  • Button/CTA remains clickable — not covered by overflowed sibling text
  • Table cell does not push neighboring columns off-screen
  • Line-clamp engages at correct max-line threshold
  • word-break: break-word present on containers that cannot use ellipsis

1.4 — Auto-Fix Rules (Deterministic)

  1. Single-line targets: overflow: hidden; text-overflow: ellipsis; white-space: nowrap;
  2. Multi-line targets: overflow: hidden; display: -webkit-box; -webkit-line-clamp: 2; -webkit-box-orient: vertical;
  3. Prose containers: word-break: break-word; overflow-wrap: anywhere;
  4. Parent with no max-width: add max-width: 100%

─── PROTOCOL 2: DATA EXTREMES INJECTION

Objective: Prove that every numeric, monetary, date, name, and identifier field renders correctly at the absolute boundaries of valid input.

2.1 — Financial / Numeric Values

ValueTypeWhat to Verify
$0ZeroDisplays as $0.00, not empty
-$1,492.50NegativeRed color token applied
$999,999,999.99MaxCell truncates, no overflow
$1,000,000,000.00Over-maxAbbreviates to $1B or ellipsis
NaNInvalidRenders —, never "NaN"
null / undefinedMissingRenders —, not "null"
InfinityOverflowCaught by validation layer
2^53Float precisionNo rounding artifacts visible

2.2 — User Identity / Name Fields

InputLengthWhat to Verify
"" (empty)0Avatar shows ?, not blank
"A"1Single char centered in avatar
"A".repeat(120)120Ellipsis + tooltip on hover
"🚀🔥💥🎯🧠"5 emojiFirst emoji or fallback shown
"🚀".repeat(40)40 emojiLine-clamped, no explosion
"'; DROP TABLE users;--"—Plain text, not executed
"<script>alert(1)</script>"—HTML-escaped, not rendered

2.3 — Date / Time Edge Cases

ValueWhat to Verify
0001-01-01No NaN or Invalid Date
9999-12-31Renders, no picker overflow
2024-02-30Normalized or clear error state
UTC +14:00Date label does not flip wrong
Unix 0 (1970-01-01)Proper date, not "Invalid Date"

2.4 — List / Table Extremes

  • 0 items: Empty state with CTA shown
  • 1 item: Singular label ("1 result" not "1 results")
  • 10,000 items: Virtualization active — no page freeze
  • 10,000 items filtered to 0: Filtered empty state (not first-use empty state)

─── PROTOCOL 3: FLAKY NETWORK STATE

Objective: Prove every network-dependent operation has defined, non-destructive behavior when the connection is degraded or lost.

3.1 — Network Failure Matrix

ScenarioTrigger PointExpected Behavior
Offline on page loadBefore first requestOffline banner + skeleton
Offline mid-navigationAfter clickOrigin preserved, retry CTA
Offline mid-form submitAfter submit clickData queued locally, queue indicator
Offline mid-streamDuring AI responsePauses at last token, auto-resumes
Timeout at 5sLong API callSpinner → timeout state → retry
Timeout at 30sVery slow endpointNever blank screen; escalation CTA
Partial JSONTruncated responseError caught → degraded state
Rapid toggle (500ms)Fast on/offDebounced — no banner flicker
Reconnect after queueBack onlineFIFO queue executes, sync shown

3.2 — Streaming-Specific Tests

  1. Cut at first token: Spinner appears before stream starts
  2. Cut mid-word: Broken character avoided; cursor/fade shown
  3. Cut at 847 chars: "Continue" or "Retry" offered — not frozen spinner
  4. No response for 10s: "Taking longer than expected... [Cancel] [Retry]"
  5. Empty stream response: Empty state renders — not a blank message bubble

3.3 — Offline Queue Verification

  • Queued count shown: "2 actions pending sync"
  • FIFO execution on reconnect
  • Duplicate-safe on rapid toggling
  • Rollback-safe: failed queue item reverts UI and shows error

Chaos Report Format

## 🎭 Chaos Report — [Page / Component]
**Date:** [ISO]   **Skill:** ux-chaos-monkey v1.0.0

### Protocol 1: Text Explosion
| Target | Bomb | Result | Auto-Fixed? |
|---|---|---|---|
| Button "Submit" | BOMB-A | ❌ Overflow 320px | ✅ overflow:hidden |
| Table: User Name | BOMB-E | ✅ Pass | — |

### Protocol 2: Data Extremes
| Field | Value | Result | Action |
|---|---|---|---|
| Price | $999,999,999.99 | ❌ Card overflow | ✅ text-overflow added |
| Username | 120-char | ✅ Ellipsis correct | — |

### Protocol 3: Flaky Network
| Scenario | Result | Action |
|---|---|---|
| Offline mid-form | ✅ Data queued | — |
| 30s timeout | ❌ White screen | 🔴 Escalate |

### Summary
- ✅ PASS: [n]   🔧 AUTO-FIXED: [n]   🔴 ESCALATE: [n]

Completion Criteria

  • All HIGH-risk text targets tested with all 5 BOMB categories
  • All financial fields tested with full numeric edge-case set
  • All date fields tested with boundary and invalid date inputs
  • All 9 network failure scenarios covered
  • Chaos Report generated with all findings documented
  • Auto-fixable issues resolved and re-tested
  • Escalations packaged with full context

Output

chaos-report-[date].md — unified 3-protocol test matrix with findings, fixes applied, and escalation packages.

Anti-patterns

  • Testing only happy-path data (clean names, round numbers, fast connections)
  • Treating null rendering as acceptable because "it rarely happens"
  • Fixing overflow with overflow: hidden alone without checking content accessibility
  • Stopping after the first failure — always run all scenarios before reporting
  • Not re-testing after auto-fixes (fix may introduce new overflow elsewhere)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.