By davidteren
Enforce intent engineering: software that behaves the way a reasonable developer or user expects. Principle- and architecture-grounded lenses for planning, plan validation, code review, and codebase audit, with project-level .intense config.
intent-engineering lens for framework architecture (Rails today). Detects structural anti-patterns (fat models, God objects, fat controllers, misused service objects, callback hell), classifies design-pattern instances against a catalog, raises unidentified patterns, and enforces the project's allow/block/approved pattern policy. Heuristic-first; optionally enriched by reek/flog/brakeman when installed.
intent-engineering lens. Reviews code and plans for convention adherence — reinvented conventions, config where convention exists, non-idiomatic structure/naming, and one-off patterns that fight the repo or framework. Reads repo CLAUDE.md/AGENTS.md FIRST; local conventions override community defaults.
intent-engineering lens for user-facing surfaces. Reviews UI/UX code and plans for missing interaction states, inconsistent look-and-feel, broken keyboard/focus/back-button, accessibility gaps, weak information architecture, and AI-slop design (HIG, look-and-feel, UX).
Always-on intent-engineering lens. Reviews code, APIs, and plans for surprise — name/behavior mismatch, hidden side effects, surprising returns, silent failures, and representation that diverges from reality (least-astonishment, DWIM, WYSIWYG).
Always-on intent-engineering lens. Reviews code and plans for needless complexity — abstractions with one use, speculative generality, knobs nobody sets, layers that don't earn their keep (Occam, KISS, YAGNI). Guards the flip side: doesn't oversimplify away real requirements.
Audit a whole codebase, subsystem, or feature against the intent-engineering lenses (predictability, convention, simplicity, experience, and architecture on supported frameworks) and produce a posture report — per-dimension 0-10 scores plus the top surprise/convention/complexity/UX/structural gaps. Use to assess an existing codebase or area, not a specific diff. Sampling-aware for large targets.
Scaffold Intent Engineering project config into a repo's .intense/ directory — ways-of-working, pattern policy (allow/block/approved), and architecture thresholds — so the team can tune the lenses and commit the settings. Stack-aware menu; idempotent (never clobbers existing config without confirmation). Use to set up or extend a project's .intense/ config.
Inject intent-engineering principles into the planning phase — surface the least-astonishment, convention, simplicity, and UX considerations relevant to the work being planned, as a tailored advisory checklist. Lightweight and non-blocking; use while drafting a plan or thinking through an approach, before a full plan doc exists. For validating a finished plan, use ie-validate-plan instead.
Review code changes through the intent-engineering lenses (predictability, convention, simplicity, experience, and architecture on supported frameworks) — surfacing surprise, non-idiomatic patterns, needless complexity, UX gaps, and structural anti-patterns. Default (interactive) mode applies safe, verified fixes and commits on a clean tree (never pushes); mode:agent reports JSON only. Use on a PR, branch, or local changes before merging.
Validate a plan, spec, or requirements document against the four intent-engineering lenses before implementation — surfacing surprising designs, non-idiomatic or reinvented approaches, needless complexity/scope, and missing UX decisions (states, flows, IA, accessibility). Returns dimensional 0-10 ratings and the gaps to resolve first. Use when a plan or spec doc exists.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Site: davidteren.github.io/intent-engineering
A Claude Code plugin that enforces intent engineering: software that behaves the way a reasonable developer or user already expects. It applies a set of well-established design principles — plus framework architecture analysis — as review lenses that fan out as parallel agents and return scored, deduplicated findings with concrete fixes.
What / Why / How (plain language)
What: A toolkit you run against a plan, a pull request, your local changes, or a whole codebase. It checks whether the work follows time-tested principles of "least surprise" — does the code/UI behave the way people expect, follow conventions, stay simple, and feel coherent?
Why: Surprising software is expensive. Hidden side effects, reinvented conventions, needless complexity, and inconsistent UX cost reviewers, users, and future maintainers time. Most of it is caught late (in review) or never. This plugin makes the principles checkable and repeatable.
How: A small set of lenses (each grounded in a researched principle) run as parallel agents. They return scored, deduplicated findings with concrete fixes, written to a report in
wip/. The same lenses work in four contexts: planning, plan validation, code review, and codebase audit.
This is the development repo for the plugin (and its marketplace). The installable plugin is self-contained under
plugins/intent-engineering/. Contributing? Read AGENTS.md first.
| Skill | Use it on | What you get |
|---|---|---|
/ie-init | a repo, once | Scaffolds .intense/ config (lens toggles, pattern policy, thresholds) so the team can tune the lenses and commit the settings. |
/ie-plan-assist | a planning draft / an approach you're weighing | An advisory checklist of the principle decisions to get right now, tailored to the work. Non-blocking. |
/ie-validate-plan | a finished plan / spec / requirements doc | Dimensional 0–10 ratings + the design gaps to resolve before coding. |
/ie-review | a PR, branch, or local changes | Findings grouped by severity; in interactive mode applies safe, verified fixes (never pushes). |
/ie-audit | a whole codebase, subsystem, or feature | A posture report — per-dimension scores and the top gaps to fix first. |
| Lens (agent) | Principles | Hunts for |
|---|---|---|
Predictability (ie-predictability-reviewer) | least-astonishment, DWIM, WYSIWYG | Name/behavior mismatch, hidden side effects, surprising returns, inconsistent branch returns, silent failures, controls that don't do what their label implies, preview/state divergence |
Convention (ie-convention-reviewer) | convention-over-configuration, framework idiom | Reinvented conventions, config where convention exists, one-off patterns that fight the repo/framework, non-idiomatic structure/naming. Reads your repo's CLAUDE.md/AGENTS.md first — local conventions win. |
Simplicity (ie-simplicity-reviewer) | Occam, KISS, YAGNI | Needless abstraction, premature generality, knobs nobody sets, layers that don't earn their keep. Also guards against over-simplifying away real requirements. |
Experience (ie-experience-reviewer) | HIG, look-and-feel, UX | Missing interaction states, inconsistent look/feel, broken keyboard/focus/back-button, accessibility gaps, weak information architecture, AI-slop design. User-facing surfaces only. |
Architecture (ie-architecture-reviewer) | structural quality, design patterns | Fat models/controllers, God objects, misused service objects, callback hell, Law of Demeter. Classifies design-pattern instances against a catalog, raises unidentified patterns, enforces your .intense/ allow/block/approved policy. Framework-specific (Rails today), code/audit only; heuristic-first, optionally enriched by reek/flog/brakeman. |
Every finding names the broken expectation (not just "this is surprising"), carries a confidence anchor, and proposes a concrete fix. When two principles conflict (e.g. DWIM's forgiving input vs least-astonishment's no-surprises), the lens flags the tension and presents the trade-off rather than dictating.
npx claudepluginhub davidteren/intent-engineering --plugin intent-engineeringComprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.