TSP-1224: Document tool swapping feature in Agent Builder prompt editor

[claude]

Summary

• prompt.mdx: Added a new "Swap tools in your prompt" subsection (positioned after "Use the Slash ( / ) Menu") explaining how to hover over a tool mention and click Swap, the two replacement sources (agent's current tools or full library), and the automatic enable/disable behavior that keeps the tool list clean. Includes a cross-reference to the Tools page.
• prompt.mdx: Updated "Reference tools in your prompt" to mention the swap feature with an anchor link to the new section.
• tools.mdx: Added a 2-sentence mention under "Adding Tools to Your Agent" about swapping tools from the prompt editor, with a cross-reference link back to the new section.
• No docs.json changes required — both pages were already in the navigation.

Linear issue: https://linear.app/relevance/issue/TSP-1224/

🤖 Generated with Claude Code

[linear]

TSP-1224

[github-actions]

🎯 Vibe check

Reviewed: 2 files (2 with issues, 0 clean)

Scores

	Dimension	Score	What's holding it back
🔴	Consistency	3/10	4 banned words (powerful ×3, robust ×1), 1 British spelling (judgement), 2 typos, 26 heading/accordion title sentence-case violations across both files
🟡	Technical clarity	7/10	Grammar error ("the model the suits"), typo in Card title ("Peformance"), and one incomplete section that reads like content was accidentally deleted
🟢	Non-technical clarity	8/10	Model comparison accordions and the Invent feature explanations are well-pitched for non-technical readers. No jargon dumps.
🟡	Structure	7/10	Redundant "Connect Tools Conveniently" section in prompt.mdx duplicates the Tools page; incomplete "Guiding" section in tools.mdx; tools.mdx lacks a closing CTA

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: The content itself is solid — the explanations are clear, the feature coverage is accurate, and the non-technical reader won't get lost. But the consistency is a significant problem: virtually every heading in both files violates sentence-case rules, and banned words turn up four times. These are mechanical fixes, but there are a lot of them.

🔧 Issues (34)

Banned words

• build/agents/build-your-agent/tools.mdx:8 — "powerful digital assistant" → remove "powerful"; e.g. "a capable digital assistant"
• build/agents/build-your-agent/tools.mdx:73 — "create a powerful assistant" → "create an assistant"
• build/agents/build-your-agent/prompt.mdx:96 — "robust function-calling abilities" → robust is banned; e.g. "strong function-calling abilities"
• build/agents/build-your-agent/prompt.mdx:101 — "provides powerful filtering" → powerful is banned; e.g. "provides filtering"

British spelling

• build/agents/build-your-agent/tools.mdx:30 — "judgement" → "judgment" (American English). Also note that line 93 already uses "judgment" correctly, making this an intra-file inconsistency.

Typos / grammar

• build/agents/build-your-agent/prompt.mdx:49 — "choose the model the suits" → "choose the model that suits"
• build/agents/build-your-agent/prompt.mdx:55 — Card title "Peformance-optimized" → "Performance-optimized"

Callout misuse

• build/agents/build-your-agent/prompt.mdx:95–97 — <Note> opens with a bold sentence acting as a label: **Not all models available in the LLM Tool step can be selected as a model for an Agent.** CLAUDE.md prohibits bold labels inside callouts. Rewrite as plain prose, or pull the bold sentence out as a proper heading above the Note.

---

Heading sentence case — prompt.mdx (16 violations)

• prompt.mdx:47 — ### Pick For Me → ### Pick for me
• prompt.mdx:60 — ### Choose Your Own Model → ### Choose your own model
• prompt.mdx:68 — Accordion "OpenAI Models" → "OpenAI models"
• prompt.mdx:73 — Accordion "Anthropic's Claude Models" → "Anthropic's Claude models"
• prompt.mdx:78 — Accordion "Azure OpenAI Models" → "Azure OpenAI models"
• prompt.mdx:83 — Accordion "Google's Gemini Models" → "Google's Gemini models"
• prompt.mdx:88 — Accordion "OpenRouter Models" → "OpenRouter models"
• prompt.mdx:99 — ### Exploring Available Models → ### Exploring available models
• prompt.mdx:104 — Accordion "Filter Options" → "Filter options"
• prompt.mdx:115 — Accordion "Sort Options" → "Sort options"
• prompt.mdx:139 — ## Prompt Instructions → ## Prompt instructions
• prompt.mdx:149 — ### Write Your Own Prompt → ### Write your own prompt
• prompt.mdx:157 — ### Use the Slash ( / ) Menu → ### Use the slash (/) menu
• prompt.mdx:187 — ## Connect Tools Conveniently → ## Connect tools conveniently
• prompt.mdx:189 — ### Add Tools to Your Agent → ### Add tools to your Agent
• prompt.mdx:197 — ## View your Prompt in Markdown → ## View your prompt in Markdown

Heading sentence case — tools.mdx (10 violations)

• tools.mdx:13 — ## Adding Tools to Your Agent → ## Adding tools to your Agent
• tools.mdx:23 — ## Tool Input Configuration Options → ## Tool input configuration options
• tools.mdx:43 — ## Guiding Your Agent's Tool Usage → ## Guiding your Agent's tool usage
• tools.mdx:53 — ## Best Practices for Tool Implementation → ## Best practices for tool implementation
• tools.mdx:63 — ## Common Tool Categories → ## Common tool categories
• tools.mdx:75 — ## Tool Approvals → ## Tool approvals
• tools.mdx:79 — ### Auto Run → ### Auto run
• tools.mdx:85 — ### Approval Required → ### Approval required
• tools.mdx:91 — ### Let Agent Decide → ### Let agent decide (already correct at line 28 — inconsistent within the same file)
• tools.mdx:97 — ## Tool Settings → ## Tool settings

🧩 Component suggestions (2)

• build/agents/build-your-agent/tools.mdx:46–51 — The "Guiding Your Agent's Tool Usage" section has only one top-level bullet (**Core Instructions**:) with nested sub-bullets under it, and nothing else. This looks like content was partially deleted — the section promises guidance on "how and when to use tools" but delivers only one item. Either restore the missing bullets (e.g. items for error handling, fallback behavior, output interpretation) or rewrite as a short prose paragraph rather than a stub list.

• build/agents/build-your-agent/prompt.mdx:95–97 — The <Note> block uses a bold opening sentence as a visual label (**Not all models available in the LLM Tool step...**). CLAUDE.md: "Callouts must be a single short paragraph — no bullet lists, no multi-line content, no bold labels inside." Options: (a) drop the bold and write it as plain prose, or (b) pull the key point out as a ### Choosing models with tool-calling support subheading above a leaner Note.

🏗️ Page structure (2)

• build/agents/build-your-agent/prompt.mdx:187–195 — The "Connect Tools Conveniently" section near the bottom repeats content already covered earlier in the same page (line 153–178) and mostly just links to the Tools page. CLAUDE.md: "No redundant sections that duplicate sidebar navigation." Cut this section; the cross-link at line 178 (For more on configuring and managing tools, see the Tools page) already does the job.

• build/agents/build-your-agent/tools.mdx (whole file) — No closing CTA. This page is part of the "Build your agent" sequential flow (as confirmed by build-overview.mdx), so a short "What's next?" section would orient the reader. Natural next steps from this repo: /build/agents/build-your-agent/triggers, /build/agents/build-your-agent/variables, or /build/agents/build-your-agent/memory.

🔋 Credit usage

Item	Count
Files reviewed	2
Context pages read	2
Total lines processed	~421

Files read: prompt.mdx (232 lines), tools.mdx (113 lines), build-overview.mdx (26 lines), variables.mdx (50 lines)