agent-security skillA
agent-security is agent-read markdown (skill) from nahid-sparktales/agent-dispatcher: Secure an agent system as a permission surface — what authority each tool call runs under, where the confused deputy sits, which controls are enforced outside the model and which are only prompt text, and how far one bad call reaches. Use when granting an agent tools or credentials, wiring in MCP servers or subagents, reviewing an agent that acted beyond what the requester could have done, or before letting an agent touch a shared or production system. Not for wording the ingestion trust boundar.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# Agent security An agent's security properties come from the tools and credentials it holds, not from how carefully its prompt is written. The design question is never "would the model refuse?" — it is "what happens when it does not?" ## When this fires An agent or subagent is being given tools, credentials or an MCP server; an agent is about to reach a shared, production or third-party system; an agent did something the person asking could not have done themselves; an agent system is being reviewed. It does not fire for a model with no tools and no outward surface. ## Procedure 1. **Write down the authority.** Every tool, every credential, every system reachable, and whose identity each call actually runs as. Most agent systems have never had this written down, and the exercise alone usually finds the problem. 2. **Find the confused deputy.** Wherever a call runs under a service account rather than the requester, the agent can be asked to do what the requester may not. Either pass the requester's own scoped identity through to the system being called, or restrict the service account to the …
Read the whole file at its exact version.
How to install
mdr add nahid-sparktales/agent-dispatcher/agent-security@git:20260919.a0d4f55mdr add nahid-sparktales/agent-dispatcher/agent-security@sha256:0cc6ee01b07f8452Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_b5d45zf2pxpywrli)
0 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (8041 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
nahid-sparktales/agent-dispatcher · 49 stars · license MIT · pushed 2026-09-23 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_b5d45zf2pxpywrli GET https://markdownregistry.com/api/v1/resolve?ref=nahid-sparktales/agent-dispatcher/agent-security GET https://markdownregistry.com/api/v1/blob/0cc6ee01b07f8452ad9f33ea6c9dd4a063e38240046100430564bf941dfcc812
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.