Skip to content

hraness/semantic-algos

v0.2.1LicenseRef-Permission-Rob-Cheung

Run explicit semantic transforms or inspectable Sem computations from Rob Cheung's work, adapted with permission.

GitHubOpen Website

Semantic Algos

Make the reasoning inspectable—only when you ask for it.

Semantic Algos packages 19 explicit semantic operators from Rob Cheung’s work as portable Agent Skills. Use a named operator for a bounded transform, or use the Sem workflow to compile, run, and present an inspectable computation over authorized text.

First proof

Ask explicitly:

Use $assumption-audit to expose the load-bearing premises in this proposal.

The operator returns an assumption register, keystone assumption, test agenda, and fallback. It does not silently route an ordinary coding or delivery request through semantic analysis.

How it works

Two interfaces share one canonical source tree:

named operator → bounded semantic transform → stated limits

request → sem-compile → inspectable program → sem-run → linked artifacts → sem-present → disposable view

The pack includes explicit transforms for assumptions, inversion, first principles, counterfactuals, abstraction, explanations, decisions, questions, analogies, why-chains, dynamic-programming decomposition, the Golden Circle, parable, joke, lyric, and Nietzsche’s Camel–Lion–Child ladder. The sem-* skills compile, execute, and present declared computations without turning them into repository orchestration.

Interfaces

  • In Codex, invoke a canonical skill explicitly as $skill-name.
  • In Cursor, invoke its thin command adapter as /skill-name.
  • Agent Skills-compatible hosts read the canonical skills/ directories.

Cursor’s commands/ adapters point back to the same skill sources and contain no duplicate procedure. cursor-explicit-skills/ intentionally suppresses automatic skill discovery so Cursor preserves the explicit-only contract.

Evidence

  • All 19 Codex skill metadata files set policy.allow_implicit_invocation: false.
  • Cursor command names map one-to-one to canonical skill directories.
  • sem-compile, sem-run, and sem-present keep checked contracts beside their skill sources.
  • Routing fixtures exclude ordinary coding, repository analysis, testing, delivery, and operations.

Pinned upstream paths and the permission record are documented in PROVENANCE.md.

Boundaries

  • Invoke a semantic operator by name; the pack must not hijack ordinary work.
  • Semantic execution may write only its declared artifacts. It does not authorize external effects.
  • Sem computation is not repository delivery, multi-agent orchestration, test execution, or host scheduling.
  • The pinned upstream revision has no published license file. This adaptation is attributed to Rob Cheung and used with permission; it is not described as MIT-licensed.

Questions

When should I use one operator instead of Sem?

Use a named operator for one bounded transform. Use sem-compile, sem-run, and optionally sem-present when the computation needs explicit dependencies, isolated execution, linked traces, or a checked view.

Why is invocation explicit?

Reasoning transforms can materially reframe a request. Explicit invocation keeps that choice with the user and prevents overlap with repository work.

Why is Cursor Marketplace distribution different?

Cursor’s public Marketplace requires a recognized permissive open-source license. The source and direct-use adapters remain here under the recorded permission, but the public Cursor listing is excluded unless the licensing gate changes.

Start

Review the named operator or Sem contract first, then install semantic-algos for authorized direct use through a supported path in the Hraness Agent Skills README.