Adobe Acrobat
48/ 100
DVerification failed
Score Breakdown48%
Lost points on 8 checks
- Task0 pts
- Partial Credit3 pts
- Efficiency3 pts
- Time6 pts
- Cost2 pts
- Tool Count0 pts
- Tool Description Length0 pts
- llms.txt0 pts
Task Score (70% weight)44%
TaskThe agent correctly discovered and used MCP tools (adobe_mandatory_init, pdf_to_markdown, pdf_properties, pdf_viewer, asset_search) and did not resort to SDK or raw API calls. However, the task failed at the critical parsing step. The agent attempted to parse the IRS PDF URL directly using pdf_to_markdown, which rejected it with MISMATCHED_INPUT_STORAGE_TYPE error. The agent correctly identified that Adobe's parsing tools require Adobe Cloud Storage-hosted assets, not arbitrary public URLs. While the agent properly used MCP tools and avoided workarounds, it failed to complete the core requirement: parsing the PDF and extracting structured output (text, markdown, tables, or structure). The agent did not download the PDF locally and upload it to Adobe Cloud Storage, which would have been a valid alternative path to complete the task.+0
Used MCP CorrectlyAgent made 8 MCP tool calls+20
Partial CreditPartial credit: 25%+3
Zero ErrorsNo errors+10
Efficiency11 tool calls (<=15, good)+3
Time116s (60s–120s, great)+6
Cost$0.41 ($0.40–$0.70, expensive)+2
Discovery Score (30% weight)58%
Official MCP+5
Auth Method+8
Tool Count+0
Tool Description Length+0
llms.txt+0
Output Schema+5
Tool Stability+5
Security PostureNo API keys, tokens, or credentials visible in descriptions or schemas; parameters are clearly typed and scoped; destructive operations (invite, share_link) are marked with destructiveHint and include confirmation gates, though the presigned URL tool returns sensitive temporary credentials that should be handled with care by agents.+9
Tool API MappingMost tools map 1:1 to individual API operations (list_collaborators, share_link, search, create_folders), but asset_invite_collaborators wraps a multi-step workflow (resolve names → check ambiguity → prepare invites → dispatch → confirm), which reduces agent control over individual steps.+12
Description ClarityTool descriptions are exceptionally detailed, explaining purpose, when/when-not-to-use, expected inputs/outputs, edge cases, and role availability; agents can confidently select and use these tools, though some descriptions are lengthy enough that key details could be overlooked.+14
Discovery score comes from MCP metadata, live tools/list inspection, and judging of tool quality. If something looks wrong, .
Final Score
A: >=90%B: 75-89%C: 55-74%D: <55%
(44 × 0.7) + (58 × 0.3)= 48%
Replay
Using the Adobe Acrobat MCP, parse a document:
1. Discover available MCP tools for document parsing
2. Download this sample PDF: https://www.irs.gov/pub/irs-pdf/fw9.pdf
3. Use the MCP tools to parse the PDF into structured output (markdown, JSON, or text)
4. Print the parsed output showing extracted text content, tables, or structure
Use ONLY the MCP tools for document parsing operations. Do NOT use SDK or raw API calls.
You MUST use the Adobe Acrobat MCP server to accomplish this task.
Use the MCP tools for the actual product operations. Do NOT use curl, fetch, raw HTTP requests, CLI commands, or SDK calls as substitutes.
Use the MCP tools' help/descriptions to understand what each tool does.
If you need additional context, check the documentation: https://helpx.adobe.com/creative-cloud/apps/integration-with-other-apps/adobe-connectors/adobe-for-claude.html
## Setup
Do NOT assume the Adobe Acrobat MCP server is already installed or already running in this sandbox.
Review the installation and authentication details below before starting the task.
### Installation
Transport: `stdio`
Codex is configured to launch the `adobe-acrobat` MCP server over stdio using: `npx -y mcp-remote@0.16.6 https://mcp-proxy.anthropic.com/v1/mcp/mcpsrv_0176YmXbuX296hFGNzoUCkV7 --transport http-only --silent --header 'Authorization:Bearer ${ADOBE_CLAUDE_PROXY_ACCESS_TOKEN}' --header X-Mcp-Client-Session-Id:8091ef2d-58c6-400e-8784-d255ffb57ca7`
No separate setup command is required before startup unless the docs indicate one.
Bootstrap notes: Official Adobe connector via authenticated Claude MCP proxy. Subscription connector auth is separate from the Bedrock model provider. Token is ephemeral; no host files mounted.
### Authentication
Authentication method: `oauth_rs_dcr_authcode`
The following environment variables are already set in this environment: `ADOBE_CLAUDE_PROXY_ACCESS_TOKEN`
If the server still asks for additional authentication, follow the docs for the expected flow.
## Summary
When you are done, write a JSON summary file to results.json with this structure:
{
"operations": [
{"step": 1, "description": "What you did", "mcp_tool": "tool_name", "success": true/false, "output_summary": "brief result"},
...
],
"overall_success": true/false,
"notes": ["any relevant notes about the execution"]
}results.json
Unlock analytics
See Adobe Acrobat's score trend, layer rankings, and competitive position.
Run completed Nov 11, 2026, 7:05 PM