executor.sh
This executive briefing evaluates the machine-readability, automated access, and tool execution readiness of executor.sh under the ARS 3.3 Standard. Using open-protocol probes across four architectural layers, Glintbase measures how effectively AI agents (such as Claude Code, Cursor, ChatGPT Sites, and LangChain agents) can discover, authenticate, and execute actions against your services without human supervision.
executor.sh is currently human-optimized. Autonomous AI agents attempting to navigate or use your service encounter substantial friction—falling back to unassisted HTML scraping and incurring severe token taxes that frequently lead to workflow timeouts.
Where autonomous agents succeed or drop off
When an AI agent executes a workflow targeting your domain, it advances through five discrete phases. If any phase lacks machine-legible declarations, the agent either halts with an exception or burns unnecessary tokens attempting heuristic recovery.
Open protocol verification results
| Protocol / Standard | Expected Vector | Scan Result | HTTP / Latency | Agent Operational Consequence |
|---|---|---|---|---|
| AI Crawler Permissions | /robots.txt | ✕ ABSENT | HTTP 200 | Controls whether ClaudeBot, GPTBot, and Perplexity index your apex pages. |
| LLMs Context Index | /llms.txt | ● VERIFIED | 404 Absent | Eliminates 90%+ token waste by providing curated markdown context to LLMs. |
| OpenAPI Specification | /openapi.json | ● VERIFIED | 404 Absent | Provides deterministic types and endpoints for agent tool generation. |
| Autonomous Machine Auth | /.well-known/auth.md | ✕ ABSENT | 404 Absent | Enables autonomous agents to register and claim API tokens without human login. |
| Model Context Protocol | /api/mcp (Streamable) | ● VERIFIED | Not Mounted | Allows Claude Desktop, Cursor, and ChatGPT Sites to run tools in your API. |
| Anti-SPA 404 Canary Guard | /_canary_404_test | ✕ ABSENT | 200 Soft-404 | Guarantees dead links return true 404 rather than HTML bundle hallucination traps. |
| WebMCP Client Tools | DOM toolname attributes | ● DEGRADED | DOM AST | Enables browser-resident agents (Chrome 157+, ChatGPT Sites) to trigger UI actions. |
| Machine Payment Protocol | HTTP 402 / MPP Header | — EXEMPT | Exempt | Allows agents to complete paid API calls and transactions autonomously. |
Position relative to industry peers
With a score of 56/100, executor.sh sits at the 75th percentile of technology organizations audited under the ARS 3.0 framework. The following table highlights the closest platform peers in your scoring band.
| Rank | Platform Domain | ARS Score | Grade | Status |
|---|---|---|---|---|
| #6 | anthropic.com | 65/100 | B | Competitive |
| #7 | resend.com | 58/100 | C | Developing |
| #8 | executor.shYOU | 56/100 | C | Developing |
| #9 | vercel.com | 38/100 | D | Developing |
| #10 | cloudflare.com | 35/100 | D | Developing |
Verified passing technical assets
Clean HTML prose (>500 chars, h1 present, valid heading progression, >5% text density).
Extracted text across core documentation stays within the 100k character budget (~25k tokens).
Standard /llms.txt file is published at apex domain or root directory.
llms.txt starts with # Heading, contains >=5 non-empty lines, valid markdown links, 100-30k chars.
Canary checks first 5 declared markdown links; all resolve to valid non-SPA content.
Server returns raw markdown payload when client sends Accept: text/markdown.
Appending .md to content and doc URLs returns raw clean markdown twins.
Verifies semantic consistency across pricing, rate limits, and authentication across documentation pages.
Critical bottlenecks impacting agent adoption
Autonomous agents operate without the contextual tolerance of humans. When critical declarations are omitted, automated tools do not guess—they abort workflows or generate hallucinated parameters. The following high-severity defects were confirmed during this audit:
Evaluates zero-shot search accuracy for multi-step agent queries (held at N/A in standard mode).
Brand-name queries rank target apex domain as the definitive authoritative origin.
Developer documentation, OpenAPI specs, and MCP endpoints are ranked in top search results.
Targeted engineering actions
public/robots.txtNo robots.txt detected; explicit AI crawler permissions recommended
$ glintbase remediate robots.txtpublic/.well-known/ard.jsonMissing /.well-known/ard.json (ARD v0.91) specification
$ glintbase remediate ard.jsonapp/not-found.tsxAnti-SPA Canary Failed: Returns soft HTTP 200 SPA shell (AI agents will hallucinate content)
$ glintbase remediate next.config.mjspublic/robots.txtEvaluates zero-shot search accuracy for multi-step agent queries (held at N/A in standard mode).
$ glintbase remediate agentic-search-usecasepublic/robots.txtBrand-name queries rank target apex domain as the definitive authoritative origin.
$ glintbase remediate brand-search-accuracyThe Glintbase Agent Readiness Score (ARS 3.0) is an open industry audit methodology calibrated against live LLM execution engines (including Claude 3.7 Sonnet, OpenAI o3, and Gemini 2.5 Flash). Probes are executed deterministically against declared endpoints. Dynamic denominators prevent archetype bias by excluding non-applicable protocol layers.