Maintainer handbook¶
← Documentation home · Using the product · Contributing · Security
This handbook covers development, system contracts, validation, and production operations. Use disposable databases with synthetic listings for development. Before production work, confirm the required approval, a verified recovery source, and exclusive writer access. If you want to browse roles or download listings, start with the user guides.
Operate from GitHub¶
For routine maintenance without a local checkout, start with routine operation from GitHub: collect, review the README proposal, deploy reviewed data separately, and verify production. The same guide identifies recovery choices and the security, dependency, and backup work that automation cannot eliminate. First-time operators must complete production activation before relying on schedules.
Develop offline¶
- Follow local setup; do not use production SQLite or enable collection.
- Read architecture and invariants before changing boundaries.
- Add a test for the behavior at risk, implement the smallest robust change, and run the affected validation gates.
- Update the relevant guide, review the complete diff, and follow the contribution workflow.
Getting started¶
| Area | Guide |
|---|---|
| Tools and first local launch | Setup |
| Paths, precedence, limits, and authorization variables | Configuration reference |
Engineering¶
| Area | Guide |
|---|---|
| Ownership, dependency direction, and generated projections | Architecture and invariants · Code-reading path |
| Visual subsystem maps and implementation links | Repository maps · Diagram source |
| Acceptance and rejection evidence | Classification |
| Discovery YAML, query identities, and request tiers | Search registry |
| Read-only queries, browser state, URL behavior, and metadata | Website engineering |
| Behavior contracts, test placement, and validation commands | Testing strategy |
| Shared writing style, navigation, generated regions, and README translations | Documentation maintenance · README localization |
| Public images, diagrams, and tour reproduction | Visual assets |
| Promotional film source recovery, editing, and replacement | Video maintenance |
Operate and recover¶
| Task | Procedure |
|---|---|
| Research overlooked LinkedIn listings and prepare reviewed additions | LinkedIn opportunity scout |
| Choose prompts for scout research, approval, and follow-up | Scout prompt templates |
| Inspect commands, inputs, side effects, and exit codes | CLI reference |
| Configure CI, schedules, protected environments, or snapshot storage | Automation |
| Review a collected state and publish it | Post-nightly runbook |
| Understand closure, timestamps, schema, or migrations | Database and lifecycle |
| Back up or restore canonical history | Database recovery |
| Build images, set permissions, or release the application | Containers and deployment |
| Roll back the served projection | Release-pointer rollback |
| Diagnose a failure without losing evidence | Operator troubleshooting |
For file-by-file inventories, use GitHub workflows, repository scripts, and test inventory. These references open on GitHub and describe individual entry points; use the guides above for step-by-step procedures.
An upstream block, failed migration, invalid snapshot, or mismatched reviewed projection is a stop condition, not permission to rebuild or bypass checks. No guide grants source-access authorization.
Releases¶
See the v1.0.0 release notes for stable product scope and compatibility. Release preparation does not authorize publication or deployment.
Agents and policy ownership¶
Coding agents start with the root agent guidelines, then load only the relevant guides. Root Security, Privacy, Contributing, Code of Conduct, and License remain canonical. Do not copy policies into new pages.
External references¶
The XERJ Corpus Hub indexes a pinned snapshot of this repository. Its selected implementation is PR #236: validated partial collections and search rotation, not the latest main branch.