Four mini-projects stack on each other over the semester, followed by an original interdisciplinary Final Project. Each mini-project must include a README and a Generative AI Use Statement (see the Policies page).
Summary of Assignments and Weights
| Assignment | Weight | Due |
|---|---|---|
| Class Participation (incl. attendance and lightning presentation) | 20% | Ongoing |
| Weekly Quizzes (lowest two dropped) | 30% | Weekly, in class |
| Mini-Project 1 — Dev Environment (Pass/No-Pass) | 5% | Fri, Sep 4 |
| Mini-Project 2 — Agent + Skills Config (Pass/No-Pass) | 5% | Fri, Sep 25 |
| Mini-Project 3 — Harness + Hooks (graded) | 10% | Fri, Oct 23 (demos Tu, Oct 27) |
| Mini-Project 4 — SDLC Capstone (graded) | 10% | Fri, Nov 20 (presentations Nov 17/19) |
| Final Project proposal (required, ungraded) | — | Fri, Nov 13 |
| Written-exam alternative opt-in deadline | — | Fri, Nov 13 |
| Final Project (poster + code + release form) | 20% | Exam period, Dec 14–18 |
| Total | 100% |
All submissions go through Moodle, with a link to a GitHub repository that is either public or privately shared with the instructor (add the instructor's GitHub account as a collaborator). Repositories must be accessible at the deadline.
Mini-Project 1 — Development Environment Configuration
5%, Pass/No-Pass. Due Friday, September 4.
Configure a professional shell/terminal environment with documented customizations (e.g., zsh + .zprofile/.zshrc, XDG base directories, a modern CLI toolkit such as ripgrep, fzf, bat, eza, tmux; VS Code with relevant extensions).
Deliverable: a GitHub dotfiles repository (public recommended for portfolio value, or privately shared) with a README documenting your configuration choices and rationale, and a .gitignore showing secrets hygiene. Graded pass/no-pass on completeness and documentation quality.
Mini-Project 2 — AI Coding Agent Configuration
5%, Pass/No-Pass. Due Friday, September 25.
High-level workflow pack. Package 3–5 reusable Claude Code customizations (Skills, a slash command, and/or a subagent) into a small project plugin a classmate can install. The point is to get repeated prompts out of chat and out of CLAUDE.md and into invocable playbooks that speed up review, status checks, and explanation.
Deliverable: the plugin repository plus a short (≤3 min) recorded or live demo showing both a typed /command and an auto-triggered Skill.
Mini-Project 3 — Agent Harness and Hooks
10%, graded. Due Friday, October 23 (demos Tuesday, October 27).
Low-level reliability harness. Add 3–5 lifecycle hooks so the agent formats, logs, injects repo state, or blocks dangerous commands whether or not the model "remembers" the rules.
Requirements: at least one hook must deny a tool call (e.g., destructive git/rm commands); at least one must produce a side effect such as lint-on-write or a Stop check that verifies tests actually ran; and the harness must integrate at least one FOSS extension from the ecosystem surveyed in Week 8 (observability/usage monitoring, context-engineering utilities, session-history tools, etc.), with the choice justified in the README.
Deliverable: the hook configuration, a trace or log showing that the runtime (not the prompt) enforced the behavior, and a 5-minute in-class demo (Week 10) walking through the deny-hook, the side-effect hook, and the trace.
Mini-Project 4 — SWE/SDLC Best Practices Capstone
10%, graded. Due Friday, November 20 (presentations Nov 17/19).
Take a real-world, high-impact, non-trivial greenfield web application (instructor-supplied, or self-scoped and instructor-approved) from ideation to a robust minimum viable product (MVP) using a complete spec-driven software development lifecycle: plan → decompose → implement → test → review, with deploy optional but strongly encouraged (e.g., GitHub Pages, Netlify, or Vercel).
Use the harness you built in MP1–3. Must include at least one instance of multi-agent/subagent orchestration (parallel dispatch, isolated worktrees, or a reviewer/implementer split) and a CI check (GitHub Actions) that runs the test suite on pull requests.
Deliverable: the repository (spec documents, PR history, CI runs, README), plus a 5-minute in-class presentation (Week 13). Graded on process adherence, final PR/deliverable quality, and documentation.
Final Project
20%. Due exam period, December 14–18.
An original interdisciplinary research project that puts the MP1–4 agent harness to work on a real research question of your choosing, applying at least one key technique from the mini-projects.
A one-page proposal (question, data/sources, planned use of the harness, expected deliverable) is due Friday, November 13 and must be approved by the instructor before substantial work begins. The deliverable is a professional research poster, code and data, a signed Kenyon Digital Release form, and publication to your public digital portfolio at digital.kenyon.edu/dh. This is the one place the course is not AI-maximalist: poster content, and especially the prose, must be your own high-touch human writing and must not read like AI slop.
Alternative Format
Students who do not wish to publish their work publicly may instead complete a 3-hour written exam in lieu of the poster/portfolio requirement, covering the semester's technical and conceptual material. Exam format and logistics will be provided separately. Students choosing this option must notify the instructor by Friday, November 13 (end of Week 12).
Final Project Requirements
- Proposal: One-page proposal due Friday, November 13, approved by the instructor before substantial work begins.
- Poster: Professional research poster in Microsoft PowerPoint format (
.ppt/.pptx), not PDF or Keynote, for digital.kenyon.edu/dh compatibility. Use the course poster template (link on the course repository). - Code: Submit code (GitHub repo, Colab notebook, or equivalent) sufficient to verify originality and reproduce results, with no secrets committed.
- Data: Submit any data required to run/test the project (e.g.,
.csvfor structured data,.pdf/documents for retrieval-based work), or document how to obtain it if it cannot be redistributed. - Poster sections: Abstract, Background (cite prior/related work), Methodology, Results, Future Work/Conclusion, Declaration of AI Use, Author Contributions (if >1 author), References (3–5, ACM format).
- External code/tutorials: If the project builds on a public tutorial, template, or framework example, cite it in references and explain in the poster how you modified, extended, or customized it (unique dataset, prompts, agents, tools, etc.).
- AI-assisted writing: Note in the Declaration of AI Use if AI helped outline, review, or generate figures for the poster. The poster prose itself must be human-written.
- Portfolio publication: Submit the signed Kenyon Digital Release Form for publication to digital.kenyon.edu/dh, or notify the instructor by Friday, November 13 to opt into the 3-hour written exam alternative instead.
- Moodle check: Confirm all grades, quizzes, and presentations are accurately recorded on Moodle.kenyon.edu before the end of the exam period.
Mini-Project 3 & 4 Grading Rubric
Each graded mini-project (MP3, MP4) is evaluated on:
| Criterion | Weight | What is assessed |
|---|---|---|
| Functionality | 15% | Harness/workflow meets the specified requirements and performs as expected |
| Technical Proficiency | 20% | Effective, non-trivial use of hooks and orchestration (MP3) or of the spec-driven process, CI, and PR workflow (MP4) |
| Documentation | 15% | Clear README/report explaining design decisions, how to reproduce the work, and the Generative AI Use Statement |
| Creativity and Extension | 25% | Thoughtful choice and integration of a FOSS extension (MP3) or scoping and specification of the application (MP4) |
| Presentation | 25% | Clarity of the in-class demo (MP3, Week 10) or presentation (MP4, Week 13), including the ability to answer questions about why the harness is designed as it is |
Pass/No-Pass mini-projects (MP1, MP2): Pass earns the full 5%. It requires a working, documented deliverable that meets all stated requirements, submitted on time. A No-Pass earns 0% until resubmitted; one resubmission within one week of feedback is allowed for up to full credit, at instructor discretion.