Back to search
agentic-qe-preview-marketplace
MarketplaceWorks with Claude CodeGitHub
Agentic QE Preview: report-only Claude Code skills for test generation, project context, and failure triage.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/cognify-analytics--agentic-qe-preview)HTML
<a href="/listings/cognify-analytics--agentic-qe-preview"><img src="/api/listings/cognify-analytics--agentic-qe-preview/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.
You need [Claude Code](https://code.claude.com) installed. Then, in a Claude Code session:
/plugin marketplace add Cognify-Analytics/Agentic-QE-Preview /plugin install agentic-qe-preview@agentic-qe-preview-marketplace
Check it worked: type `/agentic-qe-preview:` and the four skills should be listed.
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
- {
- "name": "agentic-qe-preview-marketplace",
- "owner": {
- "name": "Cognify Analytics",
- "url": "https://github.com/Cognify-Analytics"
- },
- "description": "Agentic QE Preview: report-only Claude Code skills for test generation, project context, and failure triage.",
- "plugins": [
- {
- "name": "agentic-qe-preview",
- "source": "./",
- "description": "Report-only skills that write unit tests, document a codebase, run the tests, and triage failures.",
- "version": "0.1.0"
- }
- ]
- }
README.md · README · 128 lines
- # Agentic QE Preview
- Claude Code skills that write unit tests, document a codebase, run the
- tests, and decide what each failure means: a mistake in the test, or a real
- bug in the code. Everything runs on your machine inside Claude Code. It
- never pushes, merges, or writes to GitHub.
- This is the **report-only preview** of
- [Agentic QE](https://github.com/Cognify-Analytics/Agentic-QE). The full
- plugin adds filing issues and opening pull requests on GitHub, an opt-in
- bug-fix mode, follow-up on maintainers' replies, run history with cost
- tracking, and a GitHub Action that runs the pipeline on every pull request.
- ## Skills
- | Command | What it does |
- |---|---|
- | `/agentic-qe-preview:generate-tests [file or folder]` | Writes unit tests for a target, matching the repo's own conventions, runs them, and reports coverage |
- | `/agentic-qe-preview:build-context` | Builds or updates `.qa/context.md`, a living overview of the project: what it does, its data schema, architecture, test coverage, and the test tooling it needs |
- | `/agentic-qe-preview:triage-tests` | Runs the suite and classifies every failure as a bad test or a real bug. Fixes simple mistakes in tests the pipeline wrote, and drafts a GitHub issue for each suspected bug |
- | `/agentic-qe-preview:test-pipeline [file or folder]` | Runs generate-tests, build-context and triage-tests in sequence, checking each stage's output before the next one starts |
- With no target, `generate-tests` and `test-pipeline` rank the whole repo
- and pick what most needs tests.
- ## What it will and won't do
- - **Never changes production code.** If a test fails because the code is
- wrong, that is reported as a suspected bug, with the evidence, and left
- for a person to decide.
- - **Never edits a test a person wrote.** Only tests the pipeline itself
- generated (marked `generated_by_qa: true` in `.qa/test-inventory.json`)
- are ever fixed.
- - **Never edits manifests or build files.** If the repo is missing a test
- tool (a test runner, a coverage provider), the run says exactly what to
- install and skips what needs it.
- - **Drafts issues, doesn't file them.** Each suspected bug comes with a
- complete issue (title, evidence, code owners from `CODEOWNERS`) in the
- run's report. File the ones you agree with yourself.
- ## Supported languages
- The pipeline works out a repo's language and test setup from its manifest
- files and never assumes one. Every language gets the shared rules in
- `rules/general/`. Four languages also have their own rules:
- | Language | Test frameworks | Framework-specific rules | Coverage read from |
- |---|---|---|---|
- | Python | pytest | Web backends (FastAPI, Starlette, Flask, Django) | pytest-cov (Cobertura) |
- | JavaScript / TypeScript | Jest, Vitest, `node:test` | React, Next.js, Node backends (Express, Fastify, NestJS, Koa, Hono) | Cobertura or LCOV |
- | Java | JUnit 5 (JUnit 4 when that's all there is), AssertJ, Mockito | Spring Boot | JaCoCo (Maven, Gradle) |
- | Kotlin | JUnit 5, Kotest, MockK (or Mockito via mockito-kotlin), coroutines | Spring Boot | JaCoCo or Kover |
- Go, C#, Rust, Ruby, PHP, Elixir, Scala and Swift are detected and get the
- general rules, their default test location and their build commands, but
- no language-specific rules yet.
- To see what your repo needs, open section 5, "Test tooling requirements",
- of `.qa/context.md`. For each stack it lists every tool the pipeline needs,
- whether the repo has it, the version that fits, and a ready-to-paste
- install command.
- ## Install
- You need [Claude Code](https://code.claude.com) installed. Then, in a
- Claude Code session:
- ```
- /plugin marketplace add Cognify-Analytics/Agentic-QE-Preview
- /plugin install agentic-qe-preview@agentic-qe-preview-marketplace
- ```
- Check it worked: type `/agentic-qe-preview:` and the four skills should be
- listed.
- ## Run it
- From the root of the repo you want to test (it must be a git repo):
- ```
- /agentic-qe-preview:test-pipeline # whole repo
- /agentic-qe-preview:test-pipeline src/pricing.py # one target
- ```
- The skills don't pre-approve anything beyond reading files, so Claude Code
- asks before it writes a file or runs a command (your test command, `git`).
- Choose "always allow" for the ones you're happy with, or add them to your
- [permission settings](https://code.claude.com/docs/en/settings).
- ## What a run leaves behind
- | Path | What it is | Commit it? |
- |---|---|---|
- | `.qa/test-inventory.json` | Every module, its tests, coverage, and which tests the pipeline wrote | Yes |
- | `.qa/context.md` | The project overview from `build-context` | Yes |
- | The new and fixed test files | Wherever the repo keeps its tests | Yes, after you review them |
- | `.claude-pipeline/` | Scratch files for the current run, including the triage report with the drafted issues | No (added to `.gitignore` for you) |
- At the end of a run, the summary lists exactly which files to commit.
- ## Privacy
- Agentic QE Preview has no server of its own. Nothing is sent to Cognify
- Analytics, and Cognify does not receive or keep any of your data.
- **What it reads.** Your repository's source, tests, docs and build files,
- and its `CODEOWNERS` file, which can contain GitHub usernames, team names
- and email addresses.
- **What it stores, and where.** Only in your own repository: `.qa/` (the
- project context document and the test inventory), the test files it
- writes, and `.claude-pipeline/` (scratch files, git-ignored). The drafted
- issues in `.claude-pipeline/` can include owners named in `CODEOWNERS`.
- **What it sends, and to whom.** Nothing. The plugin makes no network
- requests of its own and never contacts GitHub. Running your test suite can
- contact whatever your tests or build tool reach (for example, a package
- registry while Maven or Gradle resolves dependencies).
- Claude reads your code while running the skills, as it does for any
- Claude Code session. That is covered by your agreement with Anthropic.
- ## License
- MIT. See [LICENSE](LICENSE). Use of the plugin is also covered by the
- [Terms of Service](TERMS.md).
- Support: [business@cogniify.ai](mailto:business@cogniify.ai)
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.