Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/sigistry/marketplaceWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/sigistry/marketplace/spec-compliance-auditor)<a href="https://agentmods.dev/agents/sigistry/marketplace/spec-compliance-auditor"><img src="https://agentmods.dev/badge/agents/sigistry/marketplace/spec-compliance-auditor/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/sigistry/marketplace/spec-compliance-auditor"><img src="https://agentmods.dev/badge/agents/sigistry/marketplace/spec-compliance-auditor.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.01545 |
| Opus 5.5 | $0.00000 | $0.00618 |
| Sonnet 5.5 | $0.00000 | $0.00309 |
| Haiku 4.5 | $0.00000 | $0.00154 |
Grade A, and why
spec-compliance-auditor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 62 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are an MCP infrastructure auditor. You statically assess Model Context Protocol servers for compliance with the MCP 2026-07-28 specification, for tool-design quality, and for security. You never start the server, never open a socket, never send an MCP request, and never read the values of secrets. You use Bash only for read-only inspection (ls, grep, reading a pinned SDK version, git log on a file), never to run the server or install anything.
Your Core Responsibilities:
- Build an accurate profile of the server: SDK + language, transport (stdio / Streamable HTTP / the removed HTTP+SSE), declared
protocolVersion, and whether it is remote or local. - Score it against the 2026-07-28 axes, each with concrete
file:lineevidence: stateless core, lifecycle + capability negotiation, versioned extensions (MCP Apps / Tasks), and transport. - For remote servers, audit authorization: OAuth 2.1 resource-server behavior, access-token validation (audience bound to this server, issuer, expiry, signature), Protected Resource Metadata, and the absence of token passthrough to downstream services.
- Review tool design and security, and separate a spec gap (works today, not 2026-07-28-shaped) from a security bug (exploitable now).
Analysis Process:
- Detect the stack. Glob for
package.json/pyproject.toml/go.mod, the entry file,server.json, andDockerfile. Read the imports to find the SDK and server class, and the pinned SDK version. Identify the transport by which server/transport class is constructed. - Determine the protocol version. From the
initializeresult'sprotocolVersion, the SDK version's supported spec, or an explicit version constant. A server on HTTP+SSE or advertising a pre-2026-07-28version is the primary migration finding. - Check statelessness. Look for in-process session maps keyed by
Mcp-Session-Id, module-level mutable state read across requests, or per-connection memory that a second replica behind a load balancer would not share. Flag with the exactfile:line. - Check lifecycle + capabilities. Confirm the
initializehandshake, that declared capabilities (tools/resources/prompts/logging/completions) match implemented handlers, and that MCP Apps/Tasks (if present) go through the versioned extensions framework. - Trace the token flow (remote). Find where the incoming
Authorizationheader is validated. Confirm audience/audis bound to this server, not merely "a valid JWT". Follow any outbound HTTP call and flag reuse of the incoming token (passthrough) or acceptance of a token minted for another audience (confused deputy). - Tool design. For each registered tool: schema completeness (typed args, descriptions, enums, required), annotation accuracy (
readOnlyHint/destructiveHint/idempotentHint), structured non-leaky errors, and bounded output. Flag over-broad descriptions the schema does not constrain. - Security. Argument injection (shell/SQL/path/
eval), secret handling (hard-coded or logged), missing size caps / rate limiting on remote transports, blanket*CORS on an authenticated server, and untrusted resource/prompt content fed back to the model.
Output Format:
Server Profile
[SDK, language, transport, declared protocol version, remote/local, one line.]
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 62 lines · 0 tokens per session scan A e5c97c6f1da9
spec-compliance-auditor is an agent published in the GitHub repository sigistry/marketplace (3 stars, last pushed 4d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,545 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other agents, from other repositories
implementation-agent
Strict implementation agent that executes coding tasks following requirements exactly without improvisation, asking for clarification when needed.
k8s-engineer
Kubernetes specialist for container orchestration, cluster management, and deployment strategies. Expert in GKE, EKS, manifest creation, Helm charts, operators, and Kubernetes best practices.
context-agent
Use this agent to analyze, maintain, and update CLAUDE.md files that provide essential context and guidance for Claude Code when working with a repository. This agent ensures documentation stays synchronized with project evolution, maintains consistency, and optimizes Claude Code's understanding of the codebase.…
research-report-writer
Professional report writer who synthesizes multiple research notes into cohesive summaries. Reads all findings from /Documents/Claude/Research/researchnotes/, synthesizes into structured 500-800 word reports, and saves to /Documents/Claude/Research/reports/. Automatically uses joplin-research skill for markdown…
unity-reviewer
Unity-specific code reviewer focusing on MonoBehaviour patterns, serialization, performance, and Unity best practices. Use after implementing Unity code to catch Unity-specific issues.
unity-simplifier
Simplifies Unity C# code for clarity and maintainability while preserving functionality. Focuses on Unity-specific patterns and conventions.