Wendy Security Baselines

Organization

wendylabsinc

wendy-security-baselines

Security invariants every Wendy interop change must respect, distilled from the AAA contract. Use when: (1) reviewing or writing any code touching pki-core, wendy-auth, cloud, wendy-proxy, or wendy-agent interop, (2) designing auth, tokens, or credentials, (3) questions about the platform cap or policy ceilings, (4) SSF/CAEP events or revocation, (5) PATs, API keys, or service credentials, (6) audit logging or tamper evidence, (7) post-quantum or ML-DSA decisions, (8) TTLs, step-up, or key custody.

Overview

Publisher	wendylabsinc
Repository	`claude-skills`
Skill name	`wendy-security-baselines`
Stars	62
Forks	5
Bundled files	2
Links	Skill source Repository

Markdown instructions
A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.
Works with any LLM
AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.
2 bundled files
Scripts, templates, and references the model can read while it works. Files are read-only and never executed.
Open source
Published by wendylabsinc on GitHub. Read the source before you install it.

Installation

Install the Wendy Security Baselines AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

Open the app and go to Plugins → Skills.
Choose "Install from GitHub".
Paste the skill folder URL below and confirm.
https://github.com/wendylabsinc/claude-skills/tree/main/wendy-security-baselines
Enable the skill in any chat where you want it available.

Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills

git clone --depth 1 https://github.com/wendylabsinc/claude-skills.git /tmp/claude-skills
mkdir -p .claude/skills
cp -r /tmp/claude-skills/wendy-security-baselines .claude/skills/wendy-security-baselines

Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Wendy Security Baselines in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Wendy Security Baselines on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Wendy Security Baselines is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

Loaded only when it is needed
The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.
Switch models mid-chat
Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Wendy Security Baselines

Invariants any interop PR must respect, regardless of which side of the AAA contract it touches. This is a review checklist, not a design guide — for the responsibility split and two-key authority model, see wendy-aaa-interop; for cert issuance mechanics, wendy-certificates; for enrollment, wendy-device-enrollment; for device runtime behavior, wendy-device-interop.

The platform cap is cloud-immune

No cloud-plane path may ever touch the platform cap. It lives in pki-core configuration, owned by Wendy platform administration alone.

Every mint enforces issued ≤ min(requested, tenant ceiling, platform cap) — operator, device, and signing certs alike (§8, D4).
It is the sole bound on a compromised cloud's ability to escalate a legitimately-authenticating principal — the reason "cloud cannot authorize on its own" holds in the form that matters (§8).
If a design routes a platform-cap change through any cloud-facing API or config surface, that is a contract violation, not an optimization.

Push signals are fail-safe only

The SSF/CAEP stream carries revoke / downgrade / deprovision / session-kill — never grants.

A compromised node on this stream can over-revoke — a detectable, recoverable DoS — but cannot escalate by pushing a fake grant. Any authority increase stays on the in-band, two-key crypto path (§6, D2).
Distinguish the push stream from pki-core's on-request rights query to cloud during device-side verification: that query is cloud exercising its existing authorization authority on request, still bounded by the platform cap and gated by the principal's own signature — not a second grant path (§6).
Safe directions are all removals: auth → cloud/pki-core (deprovision, MFA revoke, group removal), cloud → pki-core (cert revoke on role change/compromise), cloud → auth (force sign-out) (§6).

No bearer secrets

No pure-bearer PATs, no copy-pasteable secrets, ever — sender-constrained tokens only.

Proof-of-possession rides as per-request signing: JWS or COSE_Sign1, algorithm ML-DSA per RFC 9964, default leaf ML-DSA-65 — never bearer-on-wire (§4.1, D17).
Explicitly not RFC 9421 HTTP Message Signatures — its registry has no ML-DSA algorithm, which would force a private, non-interoperable alg (§4.1).
Service accounts and CI get the identical operator cert flow through a different front door, never a weaker bearer-token shortcut (D17).
A cnf-bound token accepted without verifying its proof is a defect. Bound tokens go on the wire as DPoP (RFC 9449), never Bearer.

Known gaps, closing. Both are transitional, not precedent. The enforcement flip is tracked as WDY-3107, gated on the console DPoP check.

wendy-auth: "Plain-Bearer (dpop_bound: false, no cnf) is an explicit per-client opt-out, not a default." (docs/architecture.md:66)

cloud: "cnf is recorded, not enforced... until then a bound token is accepted as a bearer token" (docs/security-model.md:377-378)

Keys: client-generated, non-exportable, hardware where possible

A privileged cert is minted only behind a fresh human factor, on a key the client generated.

Human-factor-gated minting: the key alone never authenticates to wendy-auth, so a stolen key can only ride the current short-lived cert, never roll forward (§7).
Short TTLs bound the exposure window; step-up (FIDO2/passkey) is required for high-impact/destructive actions even on an otherwise-valid session (§7).
Unattended renewal is the device exception — not a general pattern — and is compensated by hardware-backed custody (TPM/OP-TEE/eFuse-DS) plus fast revocation, not by loosening the human-factor rule elsewhere (§7).

Audit is tamper-evident, secret-free, correlation-threaded

Every authority-bearing event must be recorded, attributable, tenant-scoped, and stitched by correlation_id. See references/audit.md for the full baseline and current per-service status (§12, D13).

Self-Hosted Rule (mandatory)

The wendy stack (wendyos / wendy-agent) must ALWAYS be able to run without the SaaS stack. Before shipping any interop change, consult the wendy-self-hosted skill: keep the self-hosted path intact, or create it where it doesn't exist yet. A SaaS-only feature is an incomplete feature.

Source of Truth

This skill distills the AAA Contract v0.21 (Approved 2026-08-29). The living contract lives in Linear: AAA Contract — Authentication, Authorization, Accounting. It is versioned and evolving: check it for flow-level detail, and if it disagrees with this skill, the contract wins — then update this skill.

Related Skills

wendy-aaa-interop — the responsibility split and two-key authority model these baselines enforce.
wendy-certificates — where client-generated keys and ML-DSA request-signing get applied in practice.
wendy-device-enrollment — where the hardware-custody and TTL rules land at first touch.
wendy-device-interop — where fail-safe push signals get consumed on the device side.
wendy-self-hosted — the mandatory dual-path (SaaS + self-hosted) rule.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

references/anti-patterns.md3 KB
references/audit.md3 KB

Frequently asked questions

What does the Wendy Security Baselines AI skill do?

Why use Wendy Security Baselines on TypingMind?

Because you install it once and use it with any model. Wendy Security Baselines is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Wendy Security Baselines in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/wendylabsinc/claude-skills/tree/main/wendy-security-baselines. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Wendy Security Baselines?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Wendy Security Baselines?

As many as you like. As long as a model supports skills, you can use Wendy Security Baselines with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Wendy Security Baselines AI skill free?

It is published on GitHub by wendylabsinc. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Security Triage

openclaw

OrganizationPopular

Triage OpenClaw security advisories, drafts, and GHSA reports with shipped-tag and trust-model proof.

Add Model Provider

Add and live-prove a model provider with non-interactive config one-liners, without exposing credentials.

Diagnose Gateway

Diagnose Gateway, config, secrets, channels, and port failures with read-only one-liners.

Apple Notes

Create, view, edit, delete, search, move, or export Apple Notes via the memo CLI on macOS.

Overview

Installation

Install in TypingMind

Install in another agent

Use it in TypingMind

Works with any AI model

Skill instructions

Wendy Security Baselines

The platform cap is cloud-immune

Push signals are fail-safe only

No bearer secrets

Keys: client-generated, non-exportable, hardware where possible

Audit is tamper-evident, secret-free, correlation-threaded

Self-Hosted Rule (mandatory)

Source of Truth

Related Skills

Bundled files

Frequently asked questions

Related AI Skills

Set up your own AI workspace now