Skip to content
Back to search

trace

MarketplaceWorks with Claude CodeGitHub

TRACE: automatic checking, evidence and fix verification for your coding agent.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 18 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/kg290/trace with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/kg290/trace'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/kg290--trace/badge.svg)](/listings/kg290--trace)

HTML

<a href="/listings/kg290--trace"><img src="/api/listings/kg290--trace/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

**Claude Code**, as two separate prompts:

/plugin marketplace add kg290/trace
/plugin install trace@trace

**Codex:**

codex plugin marketplace add kg290/trace
codex plugin add trace@trace

Then open `/hooks` in Codex, trust TRACE's Stop hook, and start a new thread.

**Any other agent**, or a single project without a plugin: `node install.mjs --project <path>` from a clone. Step by step, including the optional local helper: **[INSTALL.md](INSTALL.md)**.

Only install TRACE from [`kg290/trace`](https://github.com/kg290/trace) on GitHub.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 16 lines
  1. {
  2. "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
  3. "name": "trace",
  4. "description": "TRACE: automatic checking, evidence and fix verification for your coding agent.",
  5. "owner": {
  6. "name": "TRACE contributors"
  7. },
  8. "plugins": [
  9. {
  10. "name": "trace",
  11. "description": "Your agent builds it. TRACE checks it.",
  12. "source": "./",
  13. "category": "development"
  14. }
  15. ]
  16. }
README.md · README · 195 lines
  1. <p align="center">
  2. <picture>
  3. <source media="(prefers-color-scheme: dark)" srcset="assets/logo-dark.svg">
  4. <img src="assets/logo.svg" width="120" alt="TRACE logo">
  5. </picture>
  6. </p>
  7. <h1 align="center">TRACE</h1>
  8. <p align="center">
  9. <em>Your agent builds it. TRACE checks it.</em>
  10. </p>
  11. <p align="center">
  12. <a href="https://github.com/kg290/trace/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/kg290/trace/ci.yml?branch=main&style=flat-square&label=ci&color=111827" alt="CI"></a>
  13. <a href="https://github.com/kg290/trace/releases/latest"><img src="https://img.shields.io/github/v/release/kg290/trace?style=flat-square&color=111827&label=release" alt="Latest release"></a>
  14. <a href="#install"><img src="https://img.shields.io/badge/plugin-Claude%20Code%20%C2%B7%20Codex-111827?style=flat-square" alt="Plugin for Claude Code and Codex"></a>
  15. <a href="https://github.com/kg290/trace/stargazers"><img src="https://img.shields.io/github/stars/kg290/trace?style=flat-square&color=111827&label=stars" alt="Stars"></a>
  16. <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-111827?style=flat-square" alt="MIT license"></a>
  17. </p>
  18. <p align="center">
  19. <img src="assets/hero.png" width="880" alt="Three real screenshots. 1: the agent's change shows Saved, but the server answered PATCH 422. 2: TRACE reloads and the old name Alex Rivera is back, so the change was lost. 3: after the fix, the same frozen checks pass, including the 422 and 403 rejections. Baseline reproduced, 2 of 4 checks failed; after the fix, verified, 4 of 4 passed; criteria hash unchanged.">
  20. </p>
  21. <p align="center">
  22. <strong>Stop being the tester for code your AI wrote.</strong><br>
  23. TRACE runs the software, shows you the evidence, and proves the fix with the same checks that failed.<br>
  24. <sub>A plugin for Claude Code and Codex. No API keys, no telemetry. Real screenshots from the <a href="docs/DEMO.md">seeded demo</a>.</sub>
  25. </p>
  26. <p align="center">
  27. <a href="#install">Install</a> &middot;
  28. <a href="#how-it-works">How it works</a> &middot;
  29. <a href="#what-it-checks">What it checks</a> &middot;
  30. <a href="#proving-a-fix">Proving a fix</a> &middot;
  31. <a href="#faq">FAQ</a> &middot;
  32. <a href="ROADMAP.md">Roadmap</a>
  33. </p>
  34. ---
  35. ## Why
  36. You ask for a feature. The agent says it's done. You open the app, click around, find it broken, take a screenshot, paste the logs, and type *"it still doesn't work"* for the fifth time.
  37. TRACE does that part. Type `/trace` and it picks what to check from your request, the diff, the docs and the tests. It runs the software, collects the evidence, and hands your agent a compact finding instead of a wall of logs. When the agent fixes something, TRACE re-runs the **same frozen check that failed**. A pass that never failed doesn't count.
  38. ## Install
  39. **Claude Code**, as two separate prompts:
  40. ```
  41. /plugin marketplace add kg290/trace
  42. ```
  43. ```
  44. /plugin install trace@trace
  45. ```
  46. **Codex:**
  47. ```bash
  48. codex plugin marketplace add kg290/trace
  49. codex plugin add trace@trace
  50. ```
  51. Then open `/hooks` in Codex, trust TRACE's Stop hook, and start a new thread.
  52. **Any other agent**, or a single project without a plugin: `node install.mjs --project <path>` from a clone. Step by step, including the optional local helper: **[INSTALL.md](INSTALL.md)**.
  53. Only install TRACE from [`kg290/trace`](https://github.com/kg290/trace) on GitHub.
  54. ## Use it
  55. | You type | What happens |
  56. | --- | --- |
  57. | `/trace` | Checks the project, starting with what you just worked on |
  58. | `/trace check the feature you just built` | The normal path, an edge case, error handling, and the parts the change touches |
  59. | `/trace verify this fix` | Re-runs the frozen failing scenario, plus regression checks |
  60. | `/trace check the login page` | Only that area |
  61. The plugin skill lives under its plugin's name: **`/trace:trace`** in Claude Code (typing `/trace` finds it) and **`$trace:trace`** in Codex (or pick TRACE under `@`). A check-only request reports findings and changes no code. If the task includes fixing, your agent fixes and TRACE proves it.
  62. ## How it works
  63. 1. **Scope.** It reads your request, the recent diff, the README, manifests, existing tests and documented examples. Uncommitted work is left alone.
  64. 2. **Choose.** It picks at most six scenarios: the normal path, an edge or error case, the parts the change touches, and regressions. Every expectation needs a source. A guess becomes **needs review**, never a bug.
  65. 3. **Run.** It uses your agent's own browser, terminal and test tools first. An optional local helper adds exact, repeatable browser, HTTP, CLI and library checks.
  66. 4. **Report.** Each finding gives what was checked, what was expected (and from where), what actually happened, how to reproduce it, and the evidence. Duplicates are grouped and ranked by impact.
  67. 5. **Prove.** Before a fix, the checks are frozen and hashed. Afterwards, the same scenario runs again, with regression checks around it.
  68. ## What it checks
  69. | Project | What TRACE does | Evidence |
  70. | --- | --- | --- |
  71. | **Web apps** | Explores screens, fills forms, clicks, reloads, reads saved state | Screenshots, visible messages, HTTP statuses, console and runtime errors, layout measurements |
  72. | **APIs** | Calls endpoints, then reads back to confirm the write | Exact status and JSON, request and response |
  73. | **CLIs and scripts** | Runs representative commands | Exit code, stdout/stderr, generated files (a stale file can't pass) |
  74. | **Libraries** | Calls public functions with representative inputs | Return values or the expected error, async included |
  75. | **Desktop and mobile** | Only through a real driver your host provides | Otherwise **unsupported**, never faked |
  76. The helper runs Node projects directly. Python, Go, Rust and the rest go through your agent's terminal and your own test suite.
  77. ## A finding
  78. ```text
  79. ✗ CONFIRMED DEFECT · Profile change is lost after saving
  80. Checked Save "Mira Chen" as display name, then reload
  81. Expected Name persists after reload (source: examples/profile/README.md)
  82. Observed UI shows "Saved", PATCH /api/profile → 422, reload shows "Alex Rivera"
  83. Reproduce fill "Display name" → click "Save profile" → reload
  84. Evidence 3 screenshots · response status · persisted value
  85. Cause (suspected) client sends `display_name` instead of `displayName`
  86. and ignores the error response
  87. Coverage: 4 checks · 2 failed (UI + API, grouped as 1 defect) · 2 passed (422 + 403 rejections)
  88. ```
  89. Facts stay separate from suspected causes. A visual difference, a 4xx response or exit code 0 is, on its own, neither a pass nor a defect.
  90. ## Proving a fix
  91. | Result | Meaning |
  92. | --- | --- |
  93. | `reproduced` | The baseline shows the bug |
  94. | `verified fix` | The baseline failed, and the same frozen checks now pass |
  95. | `could not reproduce` | It passes, but it never failed. **Not** a fix. |
  96. | `still failing` | The fix didn't work |
  97. | `regression` | It passed before and fails now |
  98. | `blocked` · `unsupported` · `needs review` | Couldn't run, no integration, or no clear expectation |
  99. Changed criteria are refused. Every verify run keeps a screenshot for before and after, and failed checks record what was actually seen. Each result also counts the runs and the time from the failing baseline to the verified fix: measured numbers, not promises.
  100. ## Automatic checks
  101. At the end of a task that changed code, TRACE checks that change once before the agent hands back. It never runs mid-task, never asks twice in a row, and stays silent when nothing changed.
  102. | | |
  103. | --- | --- |
  104. | **Claude Code** | On with the plugin. Turn it off in `/config` → TRACE → *Automatic checks* |
  105. | **Codex** | Runs once you trust the hook in `/hooks` |
  106. | **One project off** | `{"autoCheck": false}` in that project's `.trace.json` |
  107. | **Everywhere off** | `TRACE_AUTO_CHECK=0` |
  108. ## Safe by default
  109. - **Asks first.** "Check my project" never covers payments, outside messages, destructive commands or production.
  110. - **Local only.** The helper's browser talks only to your local app. Other sites, WebSockets and service workers are blocked.
  111. - **Private stays private.** Password, token, email and card fields are masked in screenshots. Secrets are removed from text. Cookies, headers, storage and request bodies are never kept.
  112. - **Your files are safe.** Evidence lives outside your project. The installer never overwrites anything.
  113. - **Nothing leaves your machine** because of TRACE: no telemetry, no uploads, no model calls.
  114. More in [PRIVACY](docs/PRIVACY.md) and [SECURITY](SECURITY.md).
  115. ## FAQ
  116. **Does it replace my coding agent?**
  117. No. It runs inside the agent you already use. TRACE does the checking, and your agent does the fixing.
  118. **Do I need an API key or a browser download?**
  119. No. The skill uses your agent's existing tools. Only the optional helper needs Playwright's Chromium, and only for its own browser checks.
  120. **Does it save tokens?**
  121. We don't claim that. TRACE records runs and time-to-verified-fix per bug, so you can see for yourself. Checking costs tool calls too.
  122. **Can it break my data?**
  123. Read-only checks are the default. Writing test data takes an explicit permission and should use synthetic data. Project code that TRACE runs is **not** sandboxed, so only check code you trust.
  124. **If TRACE passes, is my app bug-free?**
  125. No, and TRACE never says so. It reports exactly what it covered, what passed, and what stayed untested.
  126. **Why "experimental"?**
  127. The core flows are tested on Windows with Claude Code and Codex, and CI runs Windows and Ubuntu. macOS and desktop/mobile drivers aren't covered yet. See the [validation ledger](docs/VALIDATION.md) and the [roadmap](ROADMAP.md).
  128. ## Documentation
  129. | | |
  130. | --- | --- |
  131. | [INSTALL](INSTALL.md) | Every way to install, the local helper, updating and removing |
  132. | [Demo](docs/DEMO.md) | The save that didn't save, caught and verified |
  133. | [Helper CLI](docs/CLI.md) | Commands, statuses, exit codes, permission flags |
  134. | [Architecture](docs/ARCHITECTURE.md) | How the skill, plugins, hook and helper fit together |
  135. | [Privacy](docs/PRIVACY.md) | What is collected, masked, stored and never touched |
  136. | [Validation](docs/VALIDATION.md) | What has been tested, and what hasn't |
  137. | [Design decisions](docs/RESEARCH.md) | Why TRACE is built this way |
  138. | [Roadmap](ROADMAP.md) | What's next |
  139. ## Contributing
  140. Bug reports, wrong findings and new integrations are welcome. Read [CONTRIBUTING](CONTRIBUTING.md) first, and report vulnerabilities privately ([SECURITY](SECURITY.md)). Release notes are in the [CHANGELOG](CHANGELOG.md).
  141. ## License
  142. [MIT](LICENSE). Check it, fork it, ship it.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.