skillZs
★ LIVE SKILL TAGS ★
>>> LIVE SKILLS INDEX <<<
* OPEN SOURCE *
NO LOGIN, NO TRACKING
※ REAL INSTALL DATA ※
← back to all skills
owenleezy/agent-browser-skill58 installs

agent-browser-debug

[INTERNAL TEMPLATE] Debugging execution protocol. Called by agent-browser orchestrator.

How do I install this agent skill?

npx skills add https://github.com/owenleezy/agent-browser-skill --skill agent-browser-debug
view source ↗

Is this agent skill safe to install?

  • Gen Agent Trust Hubpass

    This skill provides a structured diagnostic protocol for debugging web applications using a browser automation tool. It outlines steps to inspect JavaScript errors, network traffic, and authentication state, while explicitly instructing the agent to consult the user before logging potentially sensitive data.

  • Socketpass

    No alerts

  • Snykwarn

    Risk: MEDIUM · 1 issue

What does this agent skill do?

Debug Diagnosis — Execution Protocol

Internal template. Entry point: agent-browser skill.


NEVER

NEVER jump to DOM inspection before checking `errors` — JS errors explain 80% of broken UIs
NEVER assume auth is fine — check `cookies` + `storage local` first when login suspected
NEVER think `inspect` kills CLI — DevTools and CLI work simultaneously

Triage Sequence (run in order, stop at first finding)

agent-browser --headed open <url>
agent-browser wait --load networkidle

# Step 1: JS errors (explains 80% of broken UIs)
agent-browser errors

# Step 2: Console warnings
agent-browser console --pattern "error|warn"

# Step 3: Failed network requests — filter to relevant calls
agent-browser network requests --filter "api"
agent-browser network requests --type xhr,fetch --status 4xx   # 4xx only
agent-browser network requests --method POST --status 5xx      # failed writes

# Step 4: Inspect full request/response body of a failed request
agent-browser network request <requestId>   # get requestId from step 3

# Step 5: Auth state (check first if login/session suspected)
agent-browser cookies         # token cookie present?
agent-browser storage local   # JWT in localStorage?

# Step 6: DOM state
agent-browser snapshot -i
agent-browser is enabled @e5
agent-browser is visible @e3

# Step 7: Form validity (if button appears enabled but won't submit)
agent-browser eval "document.querySelector('form').checkValidity()"

# Step 8: Deep investigation
agent-browser inspect         # DevTools (CLI stays live)

Network HAR for Persistent Evidence

# Record full network trace (start BEFORE navigation)
agent-browser network har start ./debug-trace.har
agent-browser open <url>
agent-browser wait --load networkidle
# ... reproduce the issue ...
agent-browser network har stop
# Open debug-trace.har in browser DevTools → Network → Import

Recovery Patterns

SymptomNext action
JS errors foundReport errors → judgment gate: continue or stop?
Auth token missingRe-run auth flow, recheck cookies
Network 4xx/5xxnetwork request <id> for full body → report → judgment gate
is enabled falseCheck form validity via eval
Issue not reproducible headlessRun with --headed to observe visually

Judgment Gates

Pause when:

  • JS errors found (show to user, ask: bug or known?)
  • Auth token missing (hard to auto-recover without credentials)
  • API returning unexpected errors (user needs to decide next step)
  • network request <id> reveals sensitive data in response (don't log, ask user)

Pitfalls

MistakeFix
Skip errors, jump to DOMErrors explain 80% — always check first
Auth bug → jump to JS debuggingCheck cookies + storage local first
Think inspect kills CLIIt doesn't — both work simultaneously
Filter network requests by URL onlyAlso filter by --status 4xx or --method POST to narrow down
Miss request bodyUse network request <requestId> for full payload

Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.

<a href="https://skillzs.dev/skills/owenleezy/agent-browser-skill/agent-browser-debug">View agent-browser-debug on skillZs</a>