[KongchangAI]
Product@uri/safescript

Safescript

An open-source secure scripting language based on a subset of TypeScript with Turing-incomplete characteristics, which performs static security policy verification through compile-time data-flow graphs and taint analysis.

Timeline (last 90 days)

Nov 12

Safescript is currently a niche open-source solution, and its ecosystem maturity, available libraries, and integration with existing TypeScript toolchains all require time to be validated.

Unverified50%
Nov 12

In the interception case the compiler rejected the edit, reporting a data-flow policy violation: db:customers flowing to hooks.slack.com.

Unverified50%
Nov 12

Source-to-sink policies are declarative security rules that specify which data sources are allowed to flow to which destinations.

Unverified50%
Nov 12

Turing incompleteness results in unbounded loops being removed, and iterating over collections of unknown length requires additional abstraction.

Unverified50%
Nov 12

Safescript is an open-source project with the package name @uri/safescript, released under the MIT license.

Unverified50%
Nov 12

Safescript is a subset of TypeScript that supports JSX but removes unbounded loops, eval and dynamic host strings.

Unverified50%
Nov 12

Safescript is a Turing-incomplete language.

Unverified50%
Nov 12

Safescript can be compiled into a static data-flow graph, with security checks completed at compile time.

Unverified50%
Nov 12

In Safescript, each database query emits a namespace label (such as db:customers), each secret read emits a secret:<NAME> label, and labels propagate along the data flow.

Unverified50%
Nov 12

Safescript uses taint analysis to track the propagation of sensitive data and refuses compilation if taint flows to an unauthorized sink.

Unverified50%

5 more timeline events

All Facts (15)

Unverified

Safescript is currently a niche open-source solution, and its ecosystem maturity, available libraries, and integration with existing TypeScript toolchains all require time to be validated.

50%
Unverified

In the interception case the compiler rejected the edit, reporting a data-flow policy violation: db:customers flowing to hooks.slack.com.

50%
Unverified

Source-to-sink policies are declarative security rules that specify which data sources are allowed to flow to which destinations.

50%
Unverified

Turing incompleteness results in unbounded loops being removed, and iterating over collections of unknown length requires additional abstraction.

50%
Unverified

Safescript is an open-source project with the package name @uri/safescript, released under the MIT license.

50%
Unverified

Safescript is a subset of TypeScript that supports JSX but removes unbounded loops, eval and dynamic host strings.

50%
Unverified

Safescript is a Turing-incomplete language.

50%
Unverified

Safescript can be compiled into a static data-flow graph, with security checks completed at compile time.

50%
Unverified

In Safescript, each database query emits a namespace label (such as db:customers), each secret read emits a secret:<NAME> label, and labels propagate along the data flow.

50%
Unverified

Safescript uses taint analysis to track the propagation of sensitive data and refuses compilation if taint flows to an unauthorized sink.

50%
Unverified

The original article gives a 12-line JSX CRM example that uses SQLite tables, writing via dbTransact and reading via dbQuery.

50%
Unverified

In the interception case, the application approved the two external hosts hooks.slack.com and api.stripe.com, and the egress policy for db:customers allows only api.stripe.com.

50%
Unverified

Safescript is a restricted sandbox language with no loops, arbitrary file access, raw shell commands, or unrestricted network access.

50%
Unverified

All network requests in Safescript are statically analyzed against an explicit allowlist derived from key policies, and requests to unknown hosts are rejected before execution.

50%
Unverified

The three-layer architecture proposed by the author consists of: a first-layer sandboxed edge code (Safescript), a second-layer decision model (System 1), and a third-layer full LLM agent loop (System 2).

50%

Source Articles