Prompt Engineering

Handoff Coding Projects — Mid-Refactor State Transfer

Day three of a multi-day code change, chat nearly full: hand off the merged work, the chosen approach, the dangerous cache risk, and the cron audit still pending.

Open in Context Handoff Builder

Overview

Multi-day coding efforts carry the most intricate session state: which units are merged, which approach won (and which broke the report builder), which invariant is dangerous to miss. This setup loads a tenant-isolation marathon in Forensic fidelity — the RLS-over-rewrite decision with its reason, the analytics-replica exception as a constraint, the cross-tenant cache risk flagged as the dangerous one, the unverified-cron-paths assumption, and the audit task that must precede the cache work. The forensic rules make the next session quote this state, not reconstruct it.

How to use this resource

Carry approach plus reason

"RLS, because rewriting broke the report builder" — the why prevents relitigating the how.
Flag the dangerous risk by name

Cross-tenant cache leakage travels marked as the one to fear.
Sequence the pending work

The cron audit before the cache change — order survives the handoff.

Why This Works

Reasons traveling with decisions prevent the new session from re-deriving worse ones
Named dangers keep attention where the cost of forgetting is highest
Sequenced pending work preserves the plan, not just the list

Best for

Large refactors and migrations spanning sessions
Work with subtle invariants a fresh session would miss
Developers pairing with AI across days

Not for

Sending the source files themselves — that's the Long Prompt Splitter's codebase mode
Executing the refactor safely — that contract is the Refactor Prompt Builder's

Use cases

Continuing multi-day refactors across chats
Carrying the why behind technical choices
Keeping dangerous risks visible at the boundary

FAQ

How do the Forensic handoff rules stop a new AI chat from re-deriving decisions I already made?

The HANDOFF INSTRUCTIONS at Fidelity: Forensic tell the next session to accept everything as established context, preserve the KEY DECISIONS without silently reversing them, and quote this package rather than reconstruct from memory. So the RLS-over-query-rewriting choice travels with its reason (rewrite broke the report builder), and a fresh chat continues instead of relitigating the how.

Why does this handoff mark the cache keyspace as the dangerous risk and order the cron audit before it?

The RISKS section flags that the cache keyspace change could silently serve cross-tenant data if a key format is missed — labeled that's the dangerous one — so the highest-cost invariant stays visible. The PENDING TASKS TODO then sequences audit the cron entry points before touching the cache, preserving the plan's order across the handoff, not just the task list.

Will this handoff package carry my actual source files into the new chat, or only the session state?

It carries session state — CURRENT STATE, KEY DECISIONS, CONSTRAINTS, RISKS, and the RECOMMENDED NEXT STEP — not the code itself. notFor is explicit: sending the source files is the Long Prompt Splitter's codebase mode, and executing the refactor safely is the Refactor Prompt Builder's contract. The NOTE also reminds you items were extracted VERBATIM, so review and adjust before sending.

Customize This Resource

Opens this scenario in Context Handoff Builder. Build to get the full handoff package — then adjust the fidelity and the next step.

Open in Context Handoff Builder

Prompt Template

Copy it as-is, or use Open in Context Handoff Builder to load it pre-filled and customize it with your own context.

CONVERSATION HANDOFF PACKAGE
Paste this as the FIRST message of a new chat to continue the work. Fidelity: Forensic.

CURRENT OBJECTIVE
Land the multi-tenant isolation work across the storage layer.

CURRENT STATE
- Current state: the tenant-scoped repository wrappers are implemented for orders and invoices; both are merged and verified.

KEY DECISIONS
- Decision: we chose row-level security policies over query rewriting, after the rewrite approach broke the report builder.
- Decision: tenant context flows via the request-scoped accessor, not method parameters — that refactor is finished.

CONSTRAINTS
- Constraint: the analytics replica must keep cross-tenant reads working for the BI team; RLS must not apply there.

OPEN QUESTIONS
- Open question, not sure yet: do background jobs impersonate the tenant or run with an explicit system role?

PENDING TASKS
- Pending: the events table, the documents store, and the cache keyspace still need tenant scoping.
- TODO: audit the cron entry points for tenant context before touching the cache.
- Also still need to write the migration guide for the plugin authors.

RISKS
- Risk: the cache keyspace change could silently serve cross-tenant data if a key format is missed — that's the dangerous one.

ASSUMPTIONS
- Assumption: every entry point already passes through the middleware that sets tenant context; we verified web but not the cron paths.

RECOMMENDED NEXT STEP
Resume with: Pending: the events table, the documents store, and the cache keyspace still need tenant scoping.

HANDOFF INSTRUCTIONS
- Accept everything above as established context — do not re-derive it, and do not question it from scratch.
- Preserve the key decisions. Do not revisit or reverse them silently.
- If you believe a decision or assumption is wrong, raise it as an explicit question — never override it.
- Do not change the stated assumptions, and do not invent new ones.
- Keep tracking the open questions: resolve them or surface them, never drop them.
- Continue the pending work from the recommended next step — this is a continuation, not a fresh start.
- When referencing prior state, quote this package rather than reconstructing from memory.

NOTE
- Items above were extracted VERBATIM from the previous session (deduplicated, not paraphrased) — review and adjust before sending.

More resources from Context Handoff Builder

Resource

AI Session Handoff — Shift Change for Working Sessions

End a working session like a shift change, not an abandonment: state captured, decisions logged, next step named — ready for the next session to pick up.

Prompt Engineering

Resource

Continue a Project in a New Chat — the Handoff Package

When the window fills, don't retype the project from memory. A handoff package carries the decisions, constraints, and pending work into the new session as its first message.

Prompt Engineering

Resource

Carry Forward Open Tasks — the Work That Remains

Done work tells stories; open work needs carrying. The package separates completed from pending and resumes from the first open task.

Prompt Engineering

Resources that pair well

Resource

Estimate Token Budget — Plan Before You Paste

Token budget planning for real workloads: how much of the window a transcript actually consumes, what is left for the answer, and how much headroom remains.

Prompt Engineering

Resource

Message Too Long — the Fix That Doesn't Butcher Content

The "message too long" error has a structural fix: split at paragraph boundaries into sequenced chunks with wait rules, instead of pasting fragments and hoping.

Prompt Engineering

Resource

Convert a Conversation to a Prompt — the Full Anatomy

What a converged conversation contains: a rejected voice, an accepted framing, a hard constraint, a format decision, and a final-version moment — each mapped to its prompt section.

Prompt Engineering

Related tools

Tool

Context Handoff Builder

Continue work in a new chat without losing context — decisions, constraints, and tasks extracted into a state package.

Context Tools

Workflows that use this resource

Workflow

AI Codebase Onboarding Workflow

Get up to speed on an unfamiliar codebase in an afternoon — ground the AI in the project, have it explain the hard parts, and keep what you learn.

4 steps 30–60 minutes

Tip: Save time by exploring related resources and tools that integrate with this resource.