Documentation Project

Project Documentation Toolkit

Overview

Produce an end-to-end documentation suite for any repository by auditing the codebase, extracting technical truths, and writing markdown deliverables under docs/ plus a top-level README or overview section. Follow the workflow below to stay consistent and traceable back to source code.

Workflow Quickstart

1. Audit the repo – Load [references/researchchecklist.md](references/researchchecklist.md) and inventory languages, frameworks, configs, databases, routes, jobs, and integrations.
2. Capture hard facts – Take structured notes for runtimes, env vars, schema details, API endpoints, background jobs, deployments, and collaboration tooling.
3. Generate docs – Use the templates in [references/doctemplates.md](references/doctemplates.md) to draft each markdown file, inserting the audited facts, diagrams, and tables.
4. Cross-link and verify – Ensure docs/index.md links to every file, diagrams compile (Mermaid), tables render, and unknown data is explicitly labeled as TODO or "Not documented".
5. Final pass – Confirm folder structure, run markdown lint if available, and highlight assumptions in a final notes section if stakeholders must confirm.

Research Guidance

• Start every engagement by reading [references/researchchecklist.md](references/researchchecklist.md). It lists concrete places to inspect (migrations, routers, workers, deployment manifests) and the artifacts each doc requires.
• When facts are missing from the repo, state the gap rather than guessing, and recommend next steps (e.g., "Need PM confirmation for backup retention window").

Documentation Package

All files live under ./docs unless stated otherwise. Templates and sample snippets for each deliverable reside in [references/doctemplates.md](references/doctemplates.md).

docs/index.md (Hub + README/Overview)

• Provide the project introduction, purpose, target audience, and quick product summary.
• Include a table of contents linking to every other markdown file (relative paths) plus the README if it exists separately.
• Either create a dedicated README.md alongside docs/ or embed the overview directly in index.md. Always mention prerequisites (languages, frameworks, services) and a brief architecture recap that links to architecture.md.

docs/development.md (Development & Runtime Environment)

• Document prerequisites, installation commands, databases/services to start, and run/test scripts.
• Present env vars/config in a table (name, required, default, description) and cite where they are defined.
• Note troubleshooting steps, common migration commands, and how to seed test data.

docs/database.md

• Summarize supported databases/environments and list connection details or secrets owners.
• Draw a Mermaid ERD describing key tables, keys, and relationships (see template reference).
• For each table, describe columns with types, primary keys, indexes, and relationships. Include restore/backup steps, seed scripts, or migration commands needed locally.

docs/architecture.md

• Outline system components, deployment topology, and module boundaries.
• Include at least one Mermaid classDiagram or flowchart showing services, classes, or layers and how they interact.
• Summarize folder structure, design patterns, third-party integrations, and data flow between tiers.

docs/web-apis.md

• Catalog REST (or GraphQL) endpoints: method, route, params/query/body, auth, response formats, and error codes.
• Describe major frontend routes or views, navigation flow, and guarded pages. Provide a simple site map or flowchart when helpful.

docs/logging-and-errors.md

• Describe logging frameworks, log levels, sinks, rotation/retention, observability tooling, and correlation IDs.
• Explain error-handling approach: middleware, exception classes, retry/backoff logic, alerting, and known error codes/messages.
• Specify where logs live (files, cloud services) and how engineers monitor them.

docs/emails.md

• For every outbound email/notification, list recipients, subject, triggers, sending service, throttling rules, and template data fields.
• Include links to template files or code modules plus monitoring/alerting references.

docs/scheduled-jobs.md

• Document each cron or background job (purpose, schedule, implementation file, dependencies, datastore touchpoints, monitoring, retries).
• For ETL workloads, outline source → transform → destination plus validation/rollback procedures.

docs/operations.md (Operational Information)

• Capture environment matrix (Prod/Staging/Test/Dev) with URLs, hostnames, infra notes, secrets handling, logging destinations, and integrations/dependencies.
• Provide an access/role permission matrix, backup & retention policy summary, and detailed release plan (local build, change ticket template, release steps, rollback instructions referencing org standards if available).

docs/jira-and-git.md

• List JIRA project keys, board URLs, workflow states, ticket naming conventions, and any service desk contacts.
• Document Git remotes, default branches, branching strategy, PR requirements, CI gates, tagging/versioning, and merge responsibilities.

docs/prd.md (Feature Table)

• Build a markdown table with columns ID, Description, User Story, Expected Behavior to capture core features/backlog items.
• Use stable IDs (PRM###) and keep descriptions outcome-oriented. Add release groupings or acceptance criteria after the table if needed.

Diagram & Formatting Standards

• Mermaid blocks must render standalone in Markdown (no indentation). Validate syntax using https://mermaid.live when uncertain.
• Use sentence case headers, wrap file names in backticks (architecture.md), and link between docs for discoverability.
• Prefer declarative, audit-friendly language ("Service emits structured JSON logs" vs. "We log stuff").
• Call out assumptions or unknowns explicitly at the end of each doc section.

Validation Checklist

• docs/ folder contains: index.md, development.md, database.md, architecture.md, web-apis.md, logging-and-errors.md, emails.md, scheduled-jobs.md, operations.md, jira-and-git.md, prd.md.
• README exists (or index.md doubles as overview) and references the docs folder.
• All internal links are relative and tested.
• ERD/UML/flow diagrams reference actual schema/modules found during the audit.
• Environment tables, job schedules, and endpoint lists cite their source files where possible (e.g., "Derived from routes/api.ts").

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: ucdavis
• Source: ucdavis/ai-skills-registry
• License: MIT

Install and usage instructions live in the source repository linked above.