Harness Starter Kit Django Dogfood

harness_starter_kit_django is a dogfood repository for applying the Harness Starter Kit to a real Django project. The small server-rendered board is not primarily a product-completion exercise; it is a realistic change surface for evaluating agent rules, verification loops, decision records, failure memory, update tracking, and project-analysis quality.

The current app implements public post CRUD, search, pagination, and comments with Django models, forms, class-based views, authentication, templates, tests, and migrations.

BLOG

Features

• /: post list
• /posts/new/: create post
• /posts/<id>/: post detail
• /posts/<id>/edit/: edit post
• /posts/<id>/delete/: delete post
• /posts/<id>/comments/new/: create comment
• /comments/<id>/edit/: edit comment
• /comments/<id>/delete/: delete comment
• /accounts/signup/: sign up
• /accounts/me/: my page
• /accounts/login/: login
• /accounts/logout/: logout
• /admin/: Django admin

Post list and detail pages are public. Creating posts requires login. Updating and deleting posts is limited to the post owner. The list supports search by title, content, and owner username, plus pagination. Comments require login to create, and only the comment owner can update or delete them. New users can sign up from the web UI and are logged in automatically after registration. Post and comment create, update, delete, and validation failure flows show Django messages. Authenticated users can open a my page that summarizes their own post and comment activity and shows recent items with edit links.

Admin

The Django admin is available at /admin/. Create a superuser with:

.\.venv\Scripts\python.exe manage.py createsuperuser

Admins can manage users, groups, posts, and comments. The post admin supports owner, created-at, and updated-at filters, plus title, content, and owner search. The comment admin supports post, owner, created-at, and updated-at management.

Project Structure

.
|-- config/                         # Django project settings and root URLs
|-- harness_starter_kit_django/     # Board app
|-- .github/workflows/              # CI Harness checks
|-- .harness/                       # Harness source tracking
|-- docs/                           # Harness knowledge store
|-- evaluation/                     # Harness impact evaluation
|-- scripts/                        # Local Harness verification scripts
|-- AGENTS.md                       # Agent instructions
|-- manage.py
`-- requirements.txt

Harness Doctor Result

Harness Artifact Map

Harness Kit helped this repository create durable rules, checks, and memory rather than one-off prompts. The artifacts below are the pieces that carry that work.

flowchart TD
    A["Harness Starter Kit"] --> B["Agent instructions"]
    A --> C["Knowledge store"]
    A --> D["Drift checks"]
    A --> E["Local verification"]
    A --> F["Adoption reports"]
    A --> G["Source tracking"]
    A --> H["CI feedback"]

    B --> B1["AGENTS.md"]
    B --> B2["Project Analysis Rule"]

    C --> C1["docs/decisions/"]
    C --> C2["docs/conventions/coding.md"]
    C --> C3["docs/domain/glossary.md"]
    C --> C4["docs/failures/README.md"]

    D --> D1["scripts/check_docs_drift.py"]
    D --> D2["scripts/check_structure.py"]
    D --> D3["scripts/check_effectiveness_plan.py"]
    D --> D4["scripts/check_encoding_hygiene.py"]
    D --> D5["scripts/check_decision_memory.py"]
    D --> D6["scripts/check_failure_memory.py"]
    D --> D7[".gitattributes"]

    E --> E1["scripts/check_harness.py"]
    E --> E2["harness_starter_kit_django/tests.py"]

    F --> F1["docs/harness/adoption-report.md"]
    F --> F2["docs/harness/harness-update-report.md"]
    F --> F3["evaluation/harness-impact.md"]

    G --> G1[".harness/source.json"]
    G --> G2[".harness/decision-memory-rules.json"]

    H --> H1[".github/workflows/harness-check.yml"]

Artifact	Path	Role
Agent instructions	AGENTS.md	Project rules, commands, and forbidden actions for agents
Project analysis rule	AGENTS.md	Directs ordinary project-analysis requests through decisions, domain docs, failure notes, and Harness scripts
Adoption report	docs/harness/adoption-report.md	Harness adoption process, checks, assumptions, and remaining work
Decision records	docs/decisions/	Records choices such as Django initialization, app creation, and CRUD implementation
Coding conventions	docs/conventions/coding.md	Django structure, template, URL, and migration rules
Domain glossary	docs/domain/glossary.md	Board and Post domain vocabulary
Failure notes	docs/failures/README.md	Place to record mistakes or failed approaches that should not be repeated
Harness wrapper	scripts/check_harness.py	Runs documentation drift, structure drift, encoding hygiene, effectiveness, memory, Django checks, and tests together
Validation coverage	docs/validation.md	Binds local check scripts to the normal gate and focused validation commands
Docs drift check	scripts/check_docs_drift.py	Detects broken local references in README and docs
Structure check	scripts/check_structure.py	Detects temporary files, backup files, and drift-prone files
Encoding hygiene check	scripts/check_encoding_hygiene.py	Detects invalid UTF-8 and common Korean mojibake markers
Decision memory check	scripts/check_decision_memory.py	Warns when Django implementation diffs may need ADR coverage
Failure memory check	scripts/check_failure_memory.py	Validates that failure notes name concrete detection or prevention checks
Line ending policy	.gitattributes	Normalizes text line endings and treats images as binary
Effectiveness check	scripts/check_effectiveness_plan.py	Detects missing measurement plans in adoption and effectiveness reports
Harness source	.harness/source.json	Tracks the latest Harness Starter Kit commit referenced by this repository
Decision memory rules	.harness/decision-memory-rules.json	Configures Django paths watched by the decision-memory check
CI Harness check	.github/workflows/harness-check.yml	Runs local Harness checks in GitHub Actions
Update report	docs/harness/harness-update-report.md	Records what was applied or skipped from the latest kit update
Evaluation index	docs/evaluation.md	Separates harness health from agent effectiveness evidence
Maintenance evidence report	docs/effectiveness/effectiveness-report-harness-maintenance.md	Aggregates non-comparable Harness-maintenance outcomes
Impact evaluation	evaluation/harness-impact.md	Evaluates the practical benefits and limits of Harness Kit in this project

Harness Kit Evaluation

In this repository, Harness Kit has been more useful for collaboration rules, verification loops, decision records, failure memory, and update tracking than for the app features themselves. The detailed evaluation lives in evaluation/harness-impact.md.

Recent dogfood observation: a plain "analyze this project" request led an agent to identify structure, current behavior, test status, documentation status, git state, uncommitted Harness updates, and a 500 error path for empty comment submissions. That is a concrete example of AGENTS.md and the knowledge store raising analysis quality.

However, saying that docs/decisions/ was "read" did not always mean every decision record was read end to end. In practice it may mean search and focused summary. The latest Harness update therefore adds a Project Analysis Rule to AGENTS.md, making it more explicit that analysis requests should inspect README.md, .harness/source.json, decision records, conventions, domain docs, failure notes, and Harness scripts first.