Skip to content

The Claude Code plugin

The repository doubles as a Claude Code plugin marketplace. Installing the plugin gives Claude a skill, analysing-har, that switches on when you talk about a HAR capture, and a slash command, /analysing-har. Claude then runs WireTrail for you, installs it if it is missing, and works through the same steps you would.

Install it

In Claude Code, add the marketplace once, then install the plugin:

/plugin marketplace add johnneerdael/wiretrail
/plugin install analysing-har@analysing-har

The plugin name and the marketplace name are both analysing-har, which is why the identifier reads analysing-har@analysing-har. Check that it is installed:

/plugin

Use it

Describe the problem in plain words, or point at a file:

/analysing-har path/to/capture.har

Here is a HAR from the checkout flow. The cart does not save. What's going on?

The slash command takes the path and, optionally, a focus such as errors, auth or duplicates. The skill also switches on by itself when the conversation mentions "analyse har", "debug har" or wiretrail, when a .har file appears in the conversation or the working folder, and when you ask about request storms, duplicate calls, retries, errors, slow requests, token refresh, rate limiting, redirects, searching bodies or comparing with a baseline.

What the plugin tells Claude to do

The skill is a written set of defaults, so the agent behaves like an experienced user:

  1. Make sure WireTrail is installed, and run cargo install wiretrail if it is not.
  2. Run auto first. It prints the summary, ranks the problems and drills into the serious ones, which usually answers the whole question. auto --all adds the low-severity findings.
  3. Go deeper only where auto points, or where you named a focus, using the --filter the recommendation shows.
  4. Report the evidence: entry ids, counts, statuses and excerpts, so you can check them.
  5. Prefer --json and jq to pick out fields, and never parse the HAR JSON by hand.
  6. Use --unsafe-include-secrets only when you need a replayable request, and say so, because the output then holds live credentials.

The full text is in SKILL.md and analysing-har.md. You can read them, fork them, or paste them into another agent's instructions.

Update and uninstall

/plugin marketplace update analysing-har
/plugin update analysing-har@analysing-har

/plugin uninstall analysing-har@analysing-har
/plugin marketplace remove analysing-har

Working with an agent

Whichever agent runs wiretrail, a few habits make the result more useful and safer:

  • Give it the capture path and the question. "Why does the cart sync fail?" is answerable. "Fix my app" is not.
  • Check what it quotes. WireTrail masks the secrets it recognizes, and an agent that quotes search or extract output, or a header it does not know is secret, can repeat a credential. See How redaction works.
  • Keep captures private. A capture holds session tokens, cookies and personal data. An agent that reads one has read all of it. See Handle captures safely.
  • Ask for evidence. Entry ids let you open the same request with show-entry and check a claim.
  • Treat a finding as a lead. A ranked finding says a pattern was seen. Whether it is a bug is your decision.