docs(TSP-1233): document Has active tool failures filter

[claude]

Summary

• Adds a new Filtering tasks by error and status section to task-view.mdx explaining the Has active tool failures filter, how it works alongside Agent has errored, common use cases, and the locked-filter behavior on the Errors tab
• Adds a Filtering tasks by failure type section to task-queue.mdx with a brief introduction to both filters and a cross-reference to Task View Settings
• Adds a <Tip> callout to the Tool failures section in troubleshooting-agents-not-working.mdx pointing users to the new filter

Linear issue

https://linear.app/relevance/issue/TSP-1233/

[linear]

TSP-1233

[github-actions]

🎯 Vibe check

Reviewed: 3 files (3 with issues, 0 clean)

Scores

	Dimension	Score	What's holding it back
🔴	Consistency	5/10	Four heading-case violations in task-view.mdx, one in task-overview.mdx, and "agent"/"tool" consistently lowercased throughout troubleshooting-agents-not-working.mdx despite both being listed product-feature proper nouns.
🟡	Technical clarity	8/10	The filtering-for-failures explanation is genuinely good. Duplicate frontmatter titles across task-view.mdx and task-overview.mdx will confuse navigation and SEO; the troubleshooting page's task-overview#filtering-for-failures anchor link resolves correctly.
🟡	Non-technical clarity	7/10	task-overview.mdx opens with a good "email inbox" analogy but never defines what a task is before the layout tour. The frontmatter description promises "how tasks differ from prompts" — content that's missing from the page entirely.
🟡	Structure	7/10	Bold labels (**Examples:**, **Best Practice:**, **How it works:**) used as section sub-headings in task-view.mdx violate CLAUDE.md. Four identical "Quick Tip:" prefixes inside <Tip> components are redundant. A two-paragraph <Warning> in the troubleshooting page breaks the single-paragraph callout rule.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: The filtering content added to task-overview.mdx is the strongest part of this PR — clear, practical, and correctly cross-linked from the troubleshooting page. The troubleshooting page is thorough and well-structured, but needs a sweep to capitalize "Agent" and "Tool" throughout. task-view.mdx has the most hygiene issues: duplicate title, four heading-case violations, and bold text masquerading as headings.

🔧 Issues (8)

• build/agents/build-your-agent/agent-settings/task-view.mdx:2 — Frontmatter title: "Give Your Agent Tasks" is identical to task-overview.mdx's title. This page needs a distinct title — something like "Task view settings" — or both pages will look the same in navigation and search.

• build/agents/build-your-agent/agent-settings/task-view.mdx:11 — Heading ## Task View Settings → sentence case: ## Task view settings

• build/agents/build-your-agent/agent-settings/task-view.mdx:13 — Heading ### View Title → ### View title

• build/agents/build-your-agent/agent-settings/task-view.mdx:25 — Heading ### Which Information to Include in the View → ### Which information to include in the view

• build/agents/build-your-agent/agent-settings/task-view.mdx:37 — Heading ### When to use custom Task Views → ### When to use custom task views (Task Views is not in the capitalized product-term list; Agent, Tool, Workforce, Knowledge, Trigger are)

• build/agents/give-your-agent-tasks/task-overview.mdx:12 — Heading ## Tasks Page: Your Command Center for Agent Activity → sentence case: ## Tasks page: your command center for agent activity

• get-started/troubleshooting/troubleshooting-agents-not-working.mdx — "agent"/"agents" is lowercase throughout the entire page (lines 7, 9, 13, 37, 51, 62, 64, 65, 67, 70, 86, 90, 96, 98, 99, 119, 121, and more). Per the style guide, Agent is a capitalized product proper noun when referring to a Relevance AI Agent. Global find-and-replace on your agent→your Agent, the agent→the Agent, etc., with manual review to avoid overcapitalizing clearly generic uses.

• get-started/troubleshooting/troubleshooting-agents-not-working.mdx — "tool"/"tools" is lowercase in multiple places where it refers to the Relevance AI Tool product feature (lines 37, 51, 92, 96, 98, 100, and others). Should be "Tool"/"Tools" in product context. (The filter label "Has active tool failures" quoted verbatim from the UI is fine as-is.)

🧩 Component suggestions (6)

• build/agents/build-your-agent/agent-settings/task-view.mdx:17 — **Examples:** is bold text used as a sub-heading. CLAUDE.md explicitly says to use a proper heading instead. Replace with #### Examples.

• build/agents/build-your-agent/agent-settings/task-view.mdx:23 — **Best Practice:** Choose a name that instantly communicates... is a tip masquerading as a bold label. Convert to a <Tip> callout: <Tip>Choose a name that instantly communicates what type of tasks this view will display.</Tip>

• build/agents/build-your-agent/agent-settings/task-view.mdx:29 — **How it works:** bold label before a numbered list → use #### How it works as a proper heading.

• build/agents/give-your-agent-tasks/task-overview.mdx:34,49,64,83 — All four <Tip> callouts open with Quick Tip: — e.g. <Tip> Quick Tip: Use the search and filter options.... The component already signals its type; the prefix is redundant and adds visual clutter. Drop "Quick Tip:" from all four.

• get-started/troubleshooting/troubleshooting-agents-not-working.mdx:162–178 — "Available models and their strengths" uses three informal bullet clusters (one per provider) that repeat the same shape: strengths, best for, link. This is exactly the pattern a markdown table handles well. A three-row table with columns Model family | Strengths | Best for | Docs would let readers compare at a glance rather than scanning three separate blocks.

• get-started/troubleshooting/troubleshooting-agents-not-working.mdx:184–188 — <Warning> contains two distinct paragraphs (support-tier limitation + CTA to book a demo / Partners). CLAUDE.md says callouts must be a single short paragraph. Move the second paragraph (the book-a-demo / Partners sentence) outside the <Warning> as a regular paragraph below it, so the Warning contains only the tier-limitation statement.

🏗️ Page structure (2)

• build/agents/give-your-agent-tasks/task-overview.mdx:4 — The frontmatter description reads "What are tasks and how do they differ from prompt?" — but the page never actually explains that distinction. The content covers the Tasks page UI layout and filtering. Either update the description to match what's here ("A tour of the Tasks page: layout, sections, and how to filter by status, agent, tool, error type, and more") or add the promised concept explanation near the top.

• build/agents/give-your-agent-tasks/task-overview.mdx — The page opens immediately with the four-section workspace layout without defining what a "task" is. One or two sentences before the workspace tour would anchor non-technical readers: what constitutes a task in Relevance AI (a single run of an Agent in response to a trigger, manual prompt, or scheduled job) and why the Tasks page is where they go to review it.

⚠️ Contradictions (1)

• build/agents/give-your-agent-tasks/task-overview.mdx:32 calls the sidebar section "Saved Filters" and describes it as containing "your custom views". build/agents/build-your-agent/agent-settings/task-view.mdx:15 describes the same views as "Task Views" and says the title "will appear in the Task View selector." A reader following the task-view settings page to set up a custom view will then look for a "Task View selector" in the UI but find a section labelled "Saved Filters" instead. If these are the same feature (likely), align the terminology: pick one name and use it consistently across both pages.

🔋 Credit usage

Item	Count
Files reviewed	3
Context pages read	2
Total lines processed	~700

Files read: task-view.mdx (46 lines), task-overview.mdx (122 lines), troubleshooting-agents-not-working.mdx (199 lines), task-queue.mdx (84 lines — sibling to task-overview, checked for filtering contradictions), general.mdx (249 lines — sibling to task-view, checked for capitalization and style patterns)