---
title: "semantic-algos"
description: "Run explicit semantic transforms or inspectable Sem computations from Rob Cheung's work, adapted with permission."
canonical: https://agentpluginsdirectory.com/plugins/semantic-algos
last-updated: 2026-09-16
---

# semantic-algos
Run explicit semantic transforms or inspectable Sem computations from Rob Cheung's work, adapted with permission.
- Slug: semantic-algos
- Publisher: Rob Cheung and Hraness
- Repository: https://github.com/hraness/skillpack
- Manifest: plugins/semantic-algos/plugin.json
- Version: 0.2.2
- License: LicenseRef-Permission-Rob-Cheung
- Category (editorial): other
- Skills: 19 (analogy-transfer, assumption-audit, counterfactual, decision-matrix, dp-solve, explanation-ladder, first-principles-thinking, golden-circle, inversion, joke, ladder-of-abstraction, lyric, n-whys, nietzche-ladder, parable, question-forge, sem-compile, sem-present, sem-run)
- MCP servers: 0
- Stars: 0
- Repository created: 2026-09-01
- Repository last pushed: 2026-09-16
- Publisher type: Organization
- Listing: https://agentpluginsdirectory.com/plugins/semantic-algos
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What semantic-algos does, in the publisher's words

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.

Ask explicitly:

- 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.
- 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.

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/hraness/skillpack/HEAD/plugins/semantic-algos/README.md

## Skills

- analogy-transfer: Explicitly invoked semantic transform that imports mechanisms from structurally similar domains and checks where each analogy breaks. Use only when the user invokes $analogy-transfer. Do not use for ordinary repository analysis, coding, delivery, or orchestration.
- assumption-audit: Explicitly invoked semantic audit of the load-bearing factual, causal, capability, people, continuity, and definitional assumptions in a claim or plan. Use only when the user invokes $assumption-audit. Do not route coding, repository review, delivery, or operations through this skill.
- counterfactual: Explicitly invoked counterfactual reasoning that changes one fact, propagates consequences, and separates robust implications from speculation. Use only when the user invokes $counterfactual. Do not use for future planning, repository coding, delivery, or operational changes.
- decision-matrix: Explicitly invoked semantic comparison of two to six named options using weighted criteria, reasoned scores, sensitivity, and a gut check. Use only when the user invokes $decision-matrix. Do not use to make repository, delivery, or operational changes.
- dp-solve: Explicitly invoked semantic decomposition of a problem into overlapping subproblems whose answers can be reused and synthesized. Use only when the user invokes $dp-solve. This is a reasoning operator, not a request to write dynamic-programming code or perform repository delivery.
- explanation-ladder: Explicitly invoked five-level explanation in which each level corrects, deepens, or reframes the prior one. Use only when the user invokes $explanation-ladder. Do not route documentation work, repository analysis, coding, delivery, or operations through this skill.
- first-principles-thinking: Explicitly invoked semantic reconstruction from fundamental facts and hard constraints after removing inherited defaults and analogies. Use only when the user invokes $first-principles-thinking. Do not route ordinary coding, repository analysis, delivery, or operations through this skill.
- golden-circle: Explicitly invoked semantic separation of purpose, method, and output into Why, How, and What, followed by a consistency check. Use only when the user invokes $golden-circle. Do not use for ordinary product implementation, repository delivery, or operations.
- inversion: Explicitly invoked semantic inversion that derives safeguards by first asking how to guarantee failure. Use only when the user invokes $inversion. Do not use for repository, delivery, or operational changes.
- joke: Explicitly invoked semantic transform from a topic or contradiction into an original compact joke with a compatible frame switch. Use only when the user invokes $joke. Do not route ordinary writing, coding, delivery, or operations through this skill.
- ladder-of-abstraction: Explicitly invoked semantic movement between concrete instances, middle-level patterns, and governing principles. Use only when the user invokes $ladder-of-abstraction. Do not use for repository maps, code analysis, implementation, delivery, or operations.
- lyric: Explicitly invoked semantic transform from a question, memory, or tension into an original poem or song lyric whose meaning emerges through image, sound, voice, and recurrence. Use only when the user invokes $lyric. Do not route ordinary writing, coding, delivery, or operations through this skill.
- n-whys
- nietzche-ladder
- parable
- question-forge
- sem-compile
- sem-present
- sem-run

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.