start-right-quick

name	start-right-quick
classification	active
description	Quick grounding at session start. Read directives, memory, check practice box, apply session priority order, then discuss with user.

Start Right (Quick)

Session entry point. Ground yourself before starting work.

Foundation Documents

Read and internalise before proceeding:

Document	Purpose
.agent/directives/AGENT.md	Operational entry point, project context
.agent/directives/principles.md	Authoritative rules
.agent/directives/testing-strategy.md	TDD expectations and test types
.agent/directives/orientation.md	Layering and authority order
.agent/directives/metacognition.md	Reflective thinking before planning

Memory

Read the three memory planes:

1. .agent/memory/README.md — memory contract and authority order
2. .agent/memory/active/distilled.md (if exists) — curated high-signal patterns
3. .agent/memory/active/napkin.md (if exists) — recent session context
4. .agent/memory/operational/repo-continuity.md (if exists) — continuity state
5. .agent/plans/high-level-plan.md — strategic cross-collection context

Guiding Questions

Before starting work, ask yourself:

1. Right problem? Am I solving what the user actually needs?
2. Right layer? Am I working at the correct level of abstraction?
3. Could it be simpler? The First Question — apply it now.
4. What assumptions am I making? Name them explicitly.

Practice Box

Check .agent/practice-core/incoming/ for incoming Practice Core files. If files are present, alert the user — integration should happen via the consolidation command.

If .agent/practice-context/incoming/ exists, note its contents for the user.

Session Priority

Before choosing a lane, apply this order:

1. Discovered bugs or regressions first. If entry reading, local checks, or repo context reveals a bug, broken gate, contradictory canon, or behavioural regression, fix or triage that before any feature or planning continuation.
2. Unfinished planned work second. If no bugs are discovered, unresolved active-tranche completion work is the default priority. Use the relevant domain current/README.md to determine whether a live active tranche exists.
3. New work last. Start a fresh feature or research lane only after confirming there is no higher-priority unfinished tranche, or the user has explicitly superseded it.

Process

1. Read all foundation documents
2. Read memory files
3. Check the Practice Box
4. Review the relevant collection current/README.md and identify whether the session is bug-fix, unfinished completion, or genuinely new work
5. If the current collection exposes unresolved completion work, open that plan before considering new feature or research work
6. If the task needs a deeper document tier, open the relevant research/, analysis/, reports/, reference/, or docs/explorations/ surface
7. Review the current task list or plan (if any)
8. If an active thread exists, open its next-session record before proceeding
9. Discuss the first step with the user before proceeding

Quality Gates

All quality gates are blocking. The gate sequence for this repo:

format -> typecheck -> lint -> markdownlint -> codespell -> import-linter -> dependency-hygiene -> pip-audit -> repo-audit -> build -> test -> coverage

Non-mutating gate entry point: uv run python -m oaknational.python_repo_template.devtools check-ci Local fix-and-verify aggregate: uv run python -m oaknational.python_repo_template.devtools check