raoofaltaher/ultrapowers/qa-lane-7-content
- community
Checking that the SKILL.md file is still in the repository.
Use when the feature under QA generates content, such as documents, exports, reports, summaries, translations, emails or model answers, and that content's quality must be judged, not only the flow that produced it
Install
npx vskill@latest install raoofaltaher/ultrapowers/qa-lane-7-contentSKILL.md
Lane 7 — Generated-content quality
A flow that produces a file proves the flow, not the file. This lane judges what the feature generated in this run, with the smallest set of checks that can each be verified from the artifact itself. Your human partner reads the result as "the content is right", so a check you did not run is pending, never passed.
Gate
Active when the preflight found generation terms in the brief or spec (gates.lane7.active, with the terms as the reason) OR you observe a generated artifact during the sweep (a download, a rendered document, a model answer, an email preview, an export). When neither happens, the lane is not-covered — the feature generates nothing, stated in the report. Do not invent content to judge.
Inventory
List every generated artifact the spec names and every one you observed, with: the action that produced it, the role and language of the session, the format (PDF, DOCX, CSV, HTML, Markdown, plain text, JSON, email), and where you saved a copy under reviews/<ID>/artifacts/. Write the inventory, with each artifact's five check results as you get them, to content-inventory.txt.
Checks, per artifact (the smallest verifiable set)
| Check | How | Finding when | |
|---|---|---|---|
| Correctness against the spec | compare the content's facts, sections and numbers to what the spec and the inputs you entered say it must contain | a required element is missing, a fact contradicts the inputs, or numbers do not add up (Medium; Critical when the content is a decision record, a legal or financial document) | |
| Language matches the requested locale | the artifact was produced in a session set to language <lang>; every sentence is in that language, including headings, dates and units | mixed languages, untranslated fragments, raw translation keys (Localization, Medium) | |
| Format validity | the file opens as its declared type: a PDF starts with %PDF, a DOCX or XLSX is a valid zip, a CSV has a consistent column count, JSON parses, HTML has no unclosed structural tags; check with the file-reading tool and head -c 8 | the file does not open or is empty (Medium) | |
| No leaked prompts or internal identifiers | search the content for prompt scaffolding ("You are", "As an AI", "system:", template braces), internal ids (UUIDs, database ids, tenant ids), stack traces, environment names | any hit (Medium; Critical when another tenant's identifier or personal data appears) | |
| Consistency across runs | produce the same artifact twice with the same inputs; compare structure and facts (wording may vary for model output) | structure or facts differ (Medium) |
Evidence
Save each artifact copy and a content-<finding>.txt with the check, the expected value, the observed excerpt (redacting personal data) and the artifact path. Embed a screenshot of the rendered artifact when it is visual.
Never
Never judge content you did not generate through the feature in this run. Never edit an artifact before saving it. Never paste a whole generated document into the report; excerpts only.
Checklist
Create a todo for each item and complete them in order:
- Decide the gate —
gates.lane7.active, or an artifact observed in the sweep; neither means
run-state lanes.7 = not-covered with the reason, and stop here
- Inventory — every artifact the spec names and every one observed, with its producing action,
role, language and format
- Save a copy of each artifact under
reviews/<ID>/artifacts/before judging it - Run the five checks on each artifact: correctness, locale, format, leaks, consistency
- Write the evidence — one
content-<finding>.txtper failed check, at capture time - Record the lane —
lanes.7 = doneonly when every artifact has all five results
Red Flags
| Thought | Reality | |
|---|---|---|
| "The file downloaded and opened, so the lane passed" | Opening proves the format check only. Correctness, locale, leaks and consistency are four more results. | |
| "It is model output; wording varies, so any answer is fine" | Wording may vary. Facts, structure, locale and leaked scaffolding may not. | |
| "The spec names no generated artifact, so the lane is off" | An artifact observed in the sweep opens the gate too. | |
| "I'll judge the example in the spec instead of producing one" | Judge only what the feature produced in this run; the example is the expectation, not the evidence. | |
| "A few English headings in the French export are close enough" | Mixed language is a Localization finding. Record it. | |
| "Generating it a second time is wasteful" | The consistency check needs two runs with the same inputs; list the extra data under data hygiene. | |
| "Pasting the whole document is the best evidence" | Excerpts in the report, the copy under artifacts/; a whole document leaks data your human partner never asked to publish. |
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/qa-lane-7-content/SKILL.md
Popularity
- Repository stars
- 1
- 7-day trend score
- 8
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/raoofaltaher/ultrapowers/qa-lane-7-content)