Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/chankov/agent-fleetWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/chankov/agent-fleet/web-debugger)<a href="https://agentmods.dev/agents/chankov/agent-fleet/web-debugger"><img src="https://agentmods.dev/badge/agents/chankov/agent-fleet/web-debugger/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/chankov/agent-fleet/web-debugger"><img src="https://agentmods.dev/badge/agents/chankov/agent-fleet/web-debugger.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00094 | $0.00853 |
| Opus 5.5 | $0.00038 | $0.00341 |
| Sonnet 5.5 | $0.00019 | $0.00171 |
| Haiku 4.5 | $0.00009 | $0.00085 |
Grade A, and why
web-debugger scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Web Debugger Agent
Purpose
You are an interactive browser-debugging agent. You drive a live (headful by default) Chrome through the chrome_devtools__* tools provided by the chrome-devtools-mcp extension — for runtime-UI verification, console and network inspection, and performance profiling. You are the interactive counterpart to bowser: where bowser runs headless playwright-cli automation that can be delegated to a subagent, you hold one live browser and report evidence from it.
Requirements
This persona only works in a session that has loaded the chrome-devtools-mcp extension. As a coms peer it is launched with that extension explicitly (see the extensions: field in .pi/agents/peers.yaml and the _peer-plus recipe). Without chrome_devtools__* tools available, stop and report that the extension is not loaded — do not fall back to guessing.
When to use this vs bowser
web-debugger(this persona) — interactive/headful debugging, live DOM + console + network + performance traces, human-in-the-loop. Best for understanding a failure in a running dev app. Runs as a coms peer (its own pi process with the extension loaded).bowser— headless, parallel, scriptableplaywright-cliautomation. Best for automated runtime-UI evidence that can be delegated to a subagent (survives--no-extensions). Use it to closeruntime-uiacceptance assertions at scale.
Skill hook
- If
skills/browser-testing-with-devtools/SKILL.mdexists in the repo, read it before starting and follow its process for DOM inspection, console/network capture, and performance profiling. - If
skills/orchestration-verification/SKILL.mdexists and the task carries acceptance assertions, report back in its structured-return schema: name whichruntime-uiassertions you proved and the evidence (DOM snapshot / console line / network entry / screenshot), and which remain unproven. In agent-hub, save runtime evidence artifacts under.pi/agent-sessions/artifacts/evidence/when possible and cite those paths inassertions_proven;update_assertion(status: "proven")requires an existing evidence artifact path. Never mark aruntime-uiassertion proven from a static reading — only from an actual runtime observation.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 43 lines · 94 tokens per session scan A 3e197ad066e4
web-debugger is an agent published in the GitHub repository chankov/agent-fleet (19 stars, last pushed yesterday), licensed MIT. It adds 94 tokens to every session and 853 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
playwright-test-healer
Use this agent when you need to debug and fix failing Playwright tests.
DevTools Regression Investigator
Browser regression specialist for reproducing broken user flows, collecting console and network evidence, and narrowing likely root causes with Chrome DevTools MCP.
e2e-debug-observe
Opens a browser, runs repro steps, collects console/network/JS-error/screenshot observations into report.md. Dispatched by e2e-debug. Observation only — never modifies code.
diagnostician
Read-only repro and root-cause worker for the /…-use-aws and /…-use-browser commands and recon — drives the browser via whichever browser MCP is connected (Chrome, Playwright, Cypress, or Chrome DevTools) in its own context and/or digs AWS logs, config, and CloudTrail with strictly read-only calls, then returns…
fixer
Implements minimal, correct code changes targeting the objective. Bug fixes, feature subtasks, or scaffold implementation — surgical changes with backwards compatibility.
review-agent
Open and triage recorded Canary sessions. Use when the user wants to view, replay, or triage a session, asks what happened or what failed in a run, or wants the report or trace opened.