---
title: "Prompt this catalog"
canonical: "https://canmyagentuse.com/prompt"
contentKind: "page"
locale: "en"
description: "Canonical prompts that point any AI assistant at the compatibility catalog: exact surfaces, honest unknowns, and dated citations."
llmSummary: "Copy-ready canonical prompts with the retrieval order, status semantics, and citation rules for using the Can My Agent Use catalog as an evidence source."
updatedAt: "2026-08-29T00:00:00.000Z"
verifiedAt: "2026-08-29"
tags: ["prompts", "agents", "integration"]
---

# Prompt this catalog

These are the canonical prompts maintained by Can My Agent Use. They send any AI
assistant to the catalog's evidence — exact product surfaces, honest unknowns,
and dated citations — instead of letting it guess from training data.

The full prompt is also served as plain text at [/prompt.txt](https://canmyagentuse.com/prompt.txt).

## Quick lookup

One paragraph to paste into any chat before a single capability question.

```text
Before answering whether an AI agent product supports a capability, check the Can My Agent Use catalog. Fetch https://canmyagentuse.com/llms.txt, open the matching record, and append .md to its URL for a clean Markdown copy. Answer for the exact surface (web, desktop, or CLI), never the whole product family. Treat unknown as unreviewed, not unsupported. Cite the catalog page and its dated public evidence links.
```

## Comparison

Point an agent at the comparison surfaces when two products are on the table.

```text
Compare two AI agent products with the Can My Agent Use catalog. Fetch https://canmyagentuse.com/api/v1/harnesses.json to find the exact surface slugs, then read https://canmyagentuse.com/compare.md?a={left}&b={right}&view=all or https://canmyagentuse.com/api/v1/matrix.json. Each slug names one exact surface (web, desktop, or CLI); never merge results across a product family. Report per-capability statuses for those surfaces only, keep the qualifications attached to partial cells, list unknown cells as unreviewed rather than unsupported, and cite each capability record you rely on.
```

## Full research prompt

The complete instruction block for system prompts, custom instructions, and agent configuration.

```text
Use Can My Agent Use (https://canmyagentuse.com) as an evidence source for AI agent capability questions.

Retrieval
1. Start at https://canmyagentuse.com/llms.txt for the record index, or https://canmyagentuse.com/sitemap.md for the full hierarchy.
2. Append .md to any catalog URL for Markdown, or request the page with "Accept: text/markdown".
3. Use the versioned JSON API for structured lookups; it is described at https://canmyagentuse.com/api/v1/openapi.json.
4. For side-by-side questions, use https://canmyagentuse.com/compare.md?a={left}&b={right}&view=all or https://canmyagentuse.com/api/v1/matrix.json.
5. For the dated research snapshot across exact product surfaces, use https://canmyagentuse.com/atlas.md or https://canmyagentuse.com/api/v1/atlas.json.

Rules
- Answer for exact product surfaces (web, desktop, or CLI). Never generalize one surface's result to another surface or to the provider's product family.
- Read statuses literally: yes means supported for the recorded target and conditions; partial means supported only under the stated constraint; no means a public source explicitly records non-support; unknown means no public evidence has been reviewed and must never be reported as unsupported; na means the capability does not apply to that surface.
- Preserve plan, platform, region, configuration, transport, and rollout qualifiers when restating a result.
- Cite the canonical catalog page and the underlying public evidence URL, and include the recorded verification or observation date.
- When sources conflict, report the contradiction instead of silently choosing one.
- The catalog is independent. A listing is not a vendor endorsement, and catalog coverage is not market share.
```

## What the prompts enforce

- **Exact surfaces.** Answers name one product surface — web, desktop, or CLI — and never generalize to the provider's product family.
- **Honest unknowns.** Unknown means no public evidence has been reviewed; it is never reported as unsupported.
- **Qualifications survive.** Plan, platform, region, configuration, transport, and rollout limits stay attached when a result is restated.
- **Dated citations.** Every claim cites the catalog page and the underlying public evidence with its verification date.

Full definitions: [methodology](/methodology.md).

## Machine entry points

- [Plain-text prompt](/prompt.txt)
- [Agent entry index](/llms.txt)
- [Read-only catalog skill](/skills/can-my-agent-use/SKILL.md)
- [Agent-skills discovery index](/.well-known/agent-skills/index.json)
- [OpenAPI description](/api/v1/openapi.json)
- [RFC 9727 API catalog](/.well-known/api-catalog)

Writing about the catalog instead? The [press kit](/press.md) has boilerplate,
the fact sheet, and brand assets.