Semi Formal Reasoning
Use for code verification, debugging, and deep code analysis without running tests. This skill should be used when the user asks to "verify my changes", "check my patch", "find the bug", "why does this test fail", "are these changes correct", "review this diff", or when analyzing code semantics that require tracing execution paths. It applies semi-formal structured reasoning from the Agentic Code Reasoning paper (Ugare & Chandra, Meta, 2026) to prevent unsupported claims and force evidence-based analysis.
Install
npx skills add knot0-com/semi-formal-reasoning --skill semi-formal-reasoninggit clone --depth 1 https://github.com/knot0-com/semi-formal-reasoning.git /tmp/semi-formal-reasoning
mkdir -p ~/.claude/skills
cp -r /tmp/semi-formal-reasoning/. ~/.claude/skills/semi-formal-reasoningPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Semi-Formal Code Reasoning Structured reasoning for code verification, fault localization, and semantic analysis — without running tests. Based on the [Agentic Code Reasoning](https://arxiv.org/abs/2603.01896) paper which showed 10-12pp accuracy gains over unstructured chain-of-thought. ## Overview Standard LLM reasoning about code makes unsupported claims — assumes function behavior without tracing, skips edge cases, concludes equivalence without checking all paths. Semi-formal reasoning fixes this by requiring a structured certificate: explicit premises, execution traces with file:line citations, and formal conclusions. The agent cannot skip cases or make claims without evidence. ## When to Use - Verifying code changes are correct before committing (patch verification) - Finding the root cause of a test failure (fault localization) - Answering deep questions about code behavior (code QA) - Reviewing diffs for semantic correctness, not just style - Checking if two implementations are equivalent ## When NOT to Use - Simple typo fixes or formatting changes - When tests can be run directly (just run them) - Style/lint reviews (use a linter) ## Core Method ### Step 1: Identify the Task Type | Task | Trigger | Template | |-|-|-| | Patch Verification | "verify my changes", "check this diff", "is this correct" | Patch Equivalence | | Fault Localization | "find the bug", "why does this test fail", "where's the error" | Fault Localization |
Details
- Skill name
semi-formal-reasoning- Repository
- knot0-com/semi-formal-reasoning
- License
- MIT
- Repo updated
- 22 March 2026
FAQ
What does the Semi Formal Reasoning skill do?
Use for code verification, debugging, and deep code analysis without running tests. This skill should be used when the user asks to "verify my changes", "check my patch", "find the bug", "why does this test fail", "are these changes correct", "review this diff", or when analyzing code semantics that require tracing execution paths. It applies semi-formal structured reasoning from the Agentic Code Reasoning paper (Ugare & Chandra, Meta, 2026) to prevent unsupported claims and force evidence-based analysis.
How do I install the Semi Formal Reasoning skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add knot0-com/semi-formal-reasoning --skill semi-formal-reasoning. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Semi Formal Reasoning popular?
It lives in knot0-com/semi-formal-reasoning, which has 9 GitHub stars and 0 forks.
Related skills
Spec Driven Development
9mariano-aguero/spec-driven-development-skill
Structures feature work for AI coding agents into gated phases — research, spec, plan, tasks, implement, validate — each producing a short reviewable artifact that constrains the next. Use when requirements are informal or drift-prone, when the same prompt produces different implementations across sessions, or when a feature touches auth, database schema, or API contracts. Keywords: spec-driven development, SDD, spec.md, plan.md, tasks.md, constitution.md, acceptance criteria, drift detection.
Openclaw Genpark Vocalist
9alphaparkinc/openclaw-genpark-vocalist
OpenClaw GenPark Vocalist - Pure standard library acoustic time-domain autocorrelation pitch tracker and vocal stability estimator.
Swe CLI Skills
8SylphAI-Inc/swe-cli-skills
Senior engineer CLI expertise for AI agents — workflows, safety guardrails, gotchas, and anti-patterns across cloud, IaC, containers, databases, dev tools, and platforms
Java Development
8zander-zyx/java-development-skill
Use when writing, refactoring, or reviewing any Java or JVM code: plain Java, libraries, CLI tools, Android Java, Maven/Gradle dependency and build governance (pom.xml, BOMs, version catalogs, toolchains), Spring Boot development and version migration (2.x to 3.x, 3.x to 4.x, javax to jakarta), MyBatis/MyBatis-Plus or JPA/Hibernate persistence, writing tests with JUnit 5, Mockito, AssertJ, and Testcontainers, API design, exception handling, security review, Java upgrades, and JVM troubleshooting such as OOM, high CPU, GC tuning, GC logs, thread dumps, deadlocks, and memory leaks. Also covers J
Wechat Miniprogram Skill
9Gourdbaby/wechat-miniprogram-skill
Expert guidelines for Native WeChat Mini Program development focusing on performance, code size, and native compatibility. Use when developing WeChat Mini Programs in native JavaScript.
Golang
9saisudhir14/claude-skills
Use when writing, reviewing, or refactoring Go code. Provides production best practices for Go covering error handling, concurrency, naming, testing, performance, generics, iterators, and common pitfalls. Distilled from Google Go Style Guide, Uber Go Style Guide, Effective Go, and Go Code Review Comments. Updated for Go 1.25.
Browse the most popular Claude skills, all software development skills, or write your own with the skills generator.