Vibecode Dux-Soup

You are building a lean indie version of Dux-Soup.

Create the following project files first, then implement the application by following them. Keep the files updated as decisions change. Do not collapse this into a single README or prompt.

===== README.md =====
# Dux-Soup indie build

## Goal

Build the smallest trustworthy replacement for the core Dux-Soup workflow for one developer or a tiny team.

## Scope

Reads a CSV of LinkedIn profile URLs, drives your existing logged-in browser session to visit each one, optionally sends a connection request with a templated note, and logs outcomes to a local SQLite file with hard daily caps.

## Quick start

1. Install the documented dependencies.
2. Copy `.env.example` to `.env`.
3. Run the development command chosen during implementation.
4. Complete the acceptance checks in `BUILD_PLAN.md`.

## Honest limits

This build deliberately does not replace:
- Pacing heuristics and safety limits tuned across a lot of accounts, not guessed at by you
- Selector maintenance: when LinkedIn changes markup, someone else patches their extension and nobody patches yours
- Cloud execution, so campaigns keep running with your laptop shut
- Webhook and CRM plumbing into HubSpot, Pipedrive and friends, plus Zapier glue
- Campaign reporting, reply detection and sequence branching that actually stops when someone answers

If those capabilities are essential, use Dux-Soup instead of pretending the gap is solved.

===== AGENTS.md =====
# Agent instructions

- Optimize for a working, understandable weekend build.
- Prefer the fewest moving parts that satisfy the brief.
- Do not invent cryptography, security guarantees, APIs, or compliance claims.
- Keep secrets out of source control and logs.
- Add focused tests for destructive, security-sensitive, and data-loss paths.
- Run the project checks before declaring the build complete.
- Record any deliberate shortcut in the README under "Tradeoffs".

===== BUILD_PLAN.md =====
# Build plan

## Original build brief

Build a local LinkedIn outreach assistant that drives my own browser. Stack: TypeScript, Node 20, Playwright with a persistent Chromium user data dir, better-sqlite3 for state, no server, no cloud, no telemetry, no account system. CLI only.

Structure:
- src/browser.ts: launches Playwright chromium with userDataDir from .env (PROFILE_DIR), headed by default so I can watch and take over.
- src/login.ts: opens linkedin.com and waits for me to log in manually, then exits. Never handle or store my password. No credentials in code or .env.
- src/db.ts: SQLite schema for targets (url, name, status, added_at) and actions (target_url, action_type, result, ran_at).
- src/import.ts: reads targets.csv (column: profile_url, optional first_name) and inserts new rows as status 'queued'.
- src/run.ts: the campaign loop.

Campaign loop rules:
- Read config from config.json: daily_visit_cap, daily_connect_cap, active_hours (start and end, local time), min_delay_seconds, max_delay_seconds.
- Default caps low and boring: 20 visits, 10 connects.
- Sleep a random interval between min and max delay between every action, plus a longer random pause every 5 actions.
- Stop immediately and log if the page shows a captcha, a checkpoint, an 'unusual activity' interstitial, or a login redirect. Do not retry.
- Actions: visit profile, and optionally click Connect and add a note rendered from templates/connect.txt with {{first_name}} substitution. Truncate notes to the field limit.
- Everything runs in dry-run mode unless --live is passed. Dry run logs the exact action it would take.
- Idempotent: never act twice on the same target for the same action type.

Also include: src/report.ts printing counts by status and action outcome, a README that states plainly that this automates a logged-in session and may violate LinkedIn's user agreement and can get an account restricted, and a .env.example with PROFILE_DIR only.

Out of scope: scraping or guessing email addresses, bulk profile data export, running headless to evade detection, proxy rotation, multiple accounts, reply detection, CRM sync, any hosted component.

## Required capabilities

- Node 20+ and a local Chrome or Chromium install
- A LinkedIn account you are willing to risk, logged in via a persistent browser profile
- A CSV of target profile URLs, sourced by you
- A machine that stays awake during campaign windows
- Tolerance for re-fixing selectors whenever LinkedIn ships a UI change

## Delivery order

1. Scaffold the smallest runnable application and document its commands.
2. Implement the primary data model and core workflow.
3. Add validation, safe failure states, and persistence.
4. Cover the critical path with automated tests.
5. Exercise a clean install from the README and fix every missing step.

## Done when

- A new user can go from clone to first successful workflow using only the README.
- The core workflow works without paid infrastructure unless the brief requires it.
- Tests cover the highest-risk behavior.
- Known limitations are explicit rather than hidden.

===== .env.example =====
# Copy to .env and document every variable when it is introduced.
# Never put real credentials in this file.

APP_ENV=development
# Add only values required by the selected implementation.

You are building a lean indie version of Dux-Soup.

Create the following project files first, then implement the application by following them. Keep the files updated as decisions change. Do not collapse this into a single README or prompt.

===== README.md =====
# Dux-Soup indie build

## Goal

Build the smallest trustworthy replacement for the core Dux-Soup workflow for one developer or a tiny team.

## Scope

Reads a CSV of LinkedIn profile URLs, drives your existing logged-in browser session to visit each one, optionally sends a connection request with a templated note, and logs outcomes to a local SQLite file with hard daily caps.

## Quick start

1. Install the documented dependencies.
2. Copy `.env.example` to `.env`.
3. Run the development command chosen during implementation.
4. Complete the acceptance checks in `BUILD_PLAN.md`.

## Honest limits

This build deliberately does not replace:
- Pacing heuristics and safety limits tuned across a lot of accounts, not guessed at by you
- Selector maintenance: when LinkedIn changes markup, someone else patches their extension and nobody patches yours
- Cloud execution, so campaigns keep running with your laptop shut
- Webhook and CRM plumbing into HubSpot, Pipedrive and friends, plus Zapier glue
- Campaign reporting, reply detection and sequence branching that actually stops when someone answers

If those capabilities are essential, use Dux-Soup instead of pretending the gap is solved.

===== AGENTS.md =====
# Agent instructions

- Optimize for a working, understandable weekend build.
- Prefer the fewest moving parts that satisfy the brief.
- Do not invent cryptography, security guarantees, APIs, or compliance claims.
- Keep secrets out of source control and logs.
- Add focused tests for destructive, security-sensitive, and data-loss paths.
- Run the project checks before declaring the build complete.
- Record any deliberate shortcut in the README under "Tradeoffs".

===== BUILD_PLAN.md =====
# Build plan

## Original build brief

Build a local LinkedIn outreach assistant that drives my own browser. Stack: TypeScript, Node 20, Playwright with a persistent Chromium user data dir, better-sqlite3 for state, no server, no cloud, no telemetry, no account system. CLI only.

Structure:
- src/browser.ts: launches Playwright chromium with userDataDir from .env (PROFILE_DIR), headed by default so I can watch and take over.
- src/login.ts: opens linkedin.com and waits for me to log in manually, then exits. Never handle or store my password. No credentials in code or .env.
- src/db.ts: SQLite schema for targets (url, name, status, added_at) and actions (target_url, action_type, result, ran_at).
- src/import.ts: reads targets.csv (column: profile_url, optional first_name) and inserts new rows as status 'queued'.
- src/run.ts: the campaign loop.

Campaign loop rules:
- Read config from config.json: daily_visit_cap, daily_connect_cap, active_hours (start and end, local time), min_delay_seconds, max_delay_seconds.
- Default caps low and boring: 20 visits, 10 connects.
- Sleep a random interval between min and max delay between every action, plus a longer random pause every 5 actions.
- Stop immediately and log if the page shows a captcha, a checkpoint, an 'unusual activity' interstitial, or a login redirect. Do not retry.
- Actions: visit profile, and optionally click Connect and add a note rendered from templates/connect.txt with {{first_name}} substitution. Truncate notes to the field limit.
- Everything runs in dry-run mode unless --live is passed. Dry run logs the exact action it would take.
- Idempotent: never act twice on the same target for the same action type.

Also include: src/report.ts printing counts by status and action outcome, a README that states plainly that this automates a logged-in session and may violate LinkedIn's user agreement and can get an account restricted, and a .env.example with PROFILE_DIR only.

Out of scope: scraping or guessing email addresses, bulk profile data export, running headless to evade detection, proxy rotation, multiple accounts, reply detection, CRM sync, any hosted component.

## Required capabilities

- Node 20+ and a local Chrome or Chromium install
- A LinkedIn account you are willing to risk, logged in via a persistent browser profile
- A CSV of target profile URLs, sourced by you
- A machine that stays awake during campaign windows
- Tolerance for re-fixing selectors whenever LinkedIn ships a UI change

## Delivery order

1. Scaffold the smallest runnable application and document its commands.
2. Implement the primary data model and core workflow.
3. Add validation, safe failure states, and persistence.
4. Cover the critical path with automated tests.
5. Exercise a clean install from the README and fix every missing step.

## Done when

- A new user can go from clone to first successful workflow using only the README.
- The core workflow works without paid infrastructure unless the brief requires it.
- Tests cover the highest-risk behavior.
- Known limitations are explicit rather than hidden.

===== .env.example =====
# Copy to .env and document every variable when it is introduced.
# Never put real credentials in this file.

APP_ENV=development
# Add only values required by the selected implementation.

You are building a production product version of Dux-Soup.

Create the following project files first, then implement the application by following them. Keep the files updated as decisions change. Do not collapse this into a single README or prompt.

===== PRODUCT.md =====
# Dux-Soup product brief

## Problem

Mechanically this is a Playwright script that drives your own logged-in Chrome profile, clicks buttons, and waits a random number of seconds. An agent can get a working profile-visitor and connection-request sender running in a session, and a small campaign queue with daily caps by the end of a weekend. What you are actually buying from Dux-Soup is the accumulated caution: throttle curves that do not get your account restricted, selectors that keep working after LinkedIn reshuffles the DOM every few weeks, and a cloud mode so your laptop does not have to stay awake. Note that neither the paid tool nor your script is blessed by LinkedIn, so the ban risk is yours in both cases; you are just choosing who tunes the pacing. Build it if you want ten touches a day and enjoy fixing selectors, buy it if outreach volume is how you eat.

## Product outcome

Reads a CSV of LinkedIn profile URLs, drives your existing logged-in browser session to visit each one, optionally sends a connection request with a templated note, and logs outcomes to a local SQLite file with hard daily caps.

## Target user

A serious builder who needs a maintainable product foundation rather than a one-off demo.

## Required capabilities

- Node 20+ and a local Chrome or Chromium install
- A LinkedIn account you are willing to risk, logged in via a persistent browser profile
- A CSV of target profile URLs, sourced by you
- A machine that stays awake during campaign windows
- Tolerance for re-fixing selectors whenever LinkedIn ships a UI change

## Explicit non-goals for v1

- Pacing heuristics and safety limits tuned across a lot of accounts, not guessed at by you
- Selector maintenance: when LinkedIn changes markup, someone else patches their extension and nobody patches yours
- Cloud execution, so campaigns keep running with your laptop shut
- Webhook and CRM plumbing into HubSpot, Pipedrive and friends, plus Zapier glue
- Campaign reporting, reply detection and sequence branching that actually stops when someone answers

## Success criteria

- The primary workflow is measurable end to end.
- Setup is reproducible in a clean environment.
- Failure, recovery, and support paths are documented.
- Product claims match what the implementation actually guarantees.

===== ARCHITECTURE.md =====
# Architecture

## Starting brief

Build a local LinkedIn outreach assistant that drives my own browser. Stack: TypeScript, Node 20, Playwright with a persistent Chromium user data dir, better-sqlite3 for state, no server, no cloud, no telemetry, no account system. CLI only.

Structure:
- src/browser.ts: launches Playwright chromium with userDataDir from .env (PROFILE_DIR), headed by default so I can watch and take over.
- src/login.ts: opens linkedin.com and waits for me to log in manually, then exits. Never handle or store my password. No credentials in code or .env.
- src/db.ts: SQLite schema for targets (url, name, status, added_at) and actions (target_url, action_type, result, ran_at).
- src/import.ts: reads targets.csv (column: profile_url, optional first_name) and inserts new rows as status 'queued'.
- src/run.ts: the campaign loop.

Campaign loop rules:
- Read config from config.json: daily_visit_cap, daily_connect_cap, active_hours (start and end, local time), min_delay_seconds, max_delay_seconds.
- Default caps low and boring: 20 visits, 10 connects.
- Sleep a random interval between min and max delay between every action, plus a longer random pause every 5 actions.
- Stop immediately and log if the page shows a captcha, a checkpoint, an 'unusual activity' interstitial, or a login redirect. Do not retry.
- Actions: visit profile, and optionally click Connect and add a note rendered from templates/connect.txt with {{first_name}} substitution. Truncate notes to the field limit.
- Everything runs in dry-run mode unless --live is passed. Dry run logs the exact action it would take.
- Idempotent: never act twice on the same target for the same action type.

Also include: src/report.ts printing counts by status and action outcome, a README that states plainly that this automates a logged-in session and may violate LinkedIn's user agreement and can get an account restricted, and a .env.example with PROFILE_DIR only.

Out of scope: scraping or guessing email addresses, bulk profile data export, running headless to evade detection, proxy rotation, multiple accounts, reply detection, CRM sync, any hosted component.

## Boundaries

Separate the product into replaceable modules for interface, application logic, persistence, external integrations, and operational concerns. Keep domain logic independent from delivery frameworks and vendors.

## Production baseline

- Configuration: validated at startup with safe local defaults where possible.
- Security: least privilege, input validation, secret redaction, rate limits on abuse-prone paths, and no invented security primitives.
- Data: explicit schema and migrations, transactional writes where integrity matters, backup and restore instructions.
- Integrations: adapters around third-party providers, idempotent webhook or job processing, bounded retries, and timeouts.
- Observability: structured logs with request or operation IDs, an error-tracking hook, and health/readiness checks where a server exists.
- Quality: unit tests for domain rules, integration tests at module boundaries, and one end-to-end critical-path test.

## Decision records

For each major dependency, document why it was chosen, its failure mode, and how it can be replaced. Do not introduce infrastructure until a requirement justifies it.

===== AGENTS.md =====
# Agent instructions

- Read `PRODUCT.md` and `ARCHITECTURE.md` before changing code.
- Implement milestone by milestone; keep each change reviewable and leave the application runnable.
- Treat authentication, payments, encryption, imports, webhooks, and destructive actions as high-risk boundaries when present.
- Never invent cryptography or silently weaken a requirement to make a test pass.
- Use provider interfaces for external services and deterministic fakes in tests.
- Add migrations and rollback or recovery notes for persistent data changes.
- Log useful operational context without credentials, tokens, passwords, or personal data.
- Update documentation and run all checks before completing a milestone.

===== MILESTONES.md =====
# Delivery milestones

## M0 — Decisions and scaffold

- Confirm the runtime, persistence model, threat boundaries, and deployment target.
- Create a reproducible local environment and continuous checks.

## M1 — Core workflow

- Implement the smallest end-to-end product path with validation and tests.
- Keep integrations behind interfaces.

## M2 — Trust layer

- Add secure failure behavior, recovery paths, audit-relevant events, and data safeguards.
- Test abuse cases and destructive operations.

## M3 — Operability

- Add structured logs, error reporting hooks, health signals, backup/restore documentation, and deployment configuration.

## M4 — Release gate

- Run a clean-install test, critical-path end-to-end test, dependency review, and documented rollback exercise.
- Compare the shipped behavior with `PRODUCT.md` and publish remaining limitations.

===== OPERATIONS.md =====
# Operations

## Before release

- Validate configuration and secrets at startup.
- Define backup, restore, and rollback procedures and test them.
- Document logs, error tracking, health signals, and alert ownership.
- Set dependency update and vulnerability review expectations.

## Incident checklist

1. Contain the issue without destroying evidence or user data.
2. Record the timeline and affected scope.
3. Rotate exposed secrets and revoke compromised sessions or credentials.
4. Restore from a verified source when needed.
5. Document the root cause, remediation, and regression test.

## Launch constraint

Do not market omitted Dux-Soup capabilities as implemented. The v1 non-goals in `PRODUCT.md` remain user-visible limitations until they are deliberately delivered.

# Dux-Soup indie build

## Goal

Build the smallest trustworthy replacement for the core Dux-Soup workflow for one developer or a tiny team.

## Scope

Reads a CSV of LinkedIn profile URLs, drives your existing logged-in browser session to visit each one, optionally sends a connection request with a templated note, and logs outcomes to a local SQLite file with hard daily caps.

## Quick start

1. Install the documented dependencies.
2. Copy `.env.example` to `.env`.
3. Run the development command chosen during implementation.
4. Complete the acceptance checks in `BUILD_PLAN.md`.

## Honest limits

This build deliberately does not replace:
- Pacing heuristics and safety limits tuned across a lot of accounts, not guessed at by you
- Selector maintenance: when LinkedIn changes markup, someone else patches their extension and nobody patches yours
- Cloud execution, so campaigns keep running with your laptop shut
- Webhook and CRM plumbing into HubSpot, Pipedrive and friends, plus Zapier glue
- Campaign reporting, reply detection and sequence branching that actually stops when someone answers

If those capabilities are essential, use Dux-Soup instead of pretending the gap is solved.

# Agent instructions

- Optimize for a working, understandable weekend build.
- Prefer the fewest moving parts that satisfy the brief.
- Do not invent cryptography, security guarantees, APIs, or compliance claims.
- Keep secrets out of source control and logs.
- Add focused tests for destructive, security-sensitive, and data-loss paths.
- Run the project checks before declaring the build complete.
- Record any deliberate shortcut in the README under "Tradeoffs".

# Build plan

## Original build brief

Build a local LinkedIn outreach assistant that drives my own browser. Stack: TypeScript, Node 20, Playwright with a persistent Chromium user data dir, better-sqlite3 for state, no server, no cloud, no telemetry, no account system. CLI only.

Structure:
- src/browser.ts: launches Playwright chromium with userDataDir from .env (PROFILE_DIR), headed by default so I can watch and take over.
- src/login.ts: opens linkedin.com and waits for me to log in manually, then exits. Never handle or store my password. No credentials in code or .env.
- src/db.ts: SQLite schema for targets (url, name, status, added_at) and actions (target_url, action_type, result, ran_at).
- src/import.ts: reads targets.csv (column: profile_url, optional first_name) and inserts new rows as status 'queued'.
- src/run.ts: the campaign loop.

Campaign loop rules:
- Read config from config.json: daily_visit_cap, daily_connect_cap, active_hours (start and end, local time), min_delay_seconds, max_delay_seconds.
- Default caps low and boring: 20 visits, 10 connects.
- Sleep a random interval between min and max delay between every action, plus a longer random pause every 5 actions.
- Stop immediately and log if the page shows a captcha, a checkpoint, an 'unusual activity' interstitial, or a login redirect. Do not retry.
- Actions: visit profile, and optionally click Connect and add a note rendered from templates/connect.txt with {{first_name}} substitution. Truncate notes to the field limit.
- Everything runs in dry-run mode unless --live is passed. Dry run logs the exact action it would take.
- Idempotent: never act twice on the same target for the same action type.

Also include: src/report.ts printing counts by status and action outcome, a README that states plainly that this automates a logged-in session and may violate LinkedIn's user agreement and can get an account restricted, and a .env.example with PROFILE_DIR only.

Out of scope: scraping or guessing email addresses, bulk profile data export, running headless to evade detection, proxy rotation, multiple accounts, reply detection, CRM sync, any hosted component.

## Required capabilities

- Node 20+ and a local Chrome or Chromium install
- A LinkedIn account you are willing to risk, logged in via a persistent browser profile
- A CSV of target profile URLs, sourced by you
- A machine that stays awake during campaign windows
- Tolerance for re-fixing selectors whenever LinkedIn ships a UI change

## Delivery order

1. Scaffold the smallest runnable application and document its commands.
2. Implement the primary data model and core workflow.
3. Add validation, safe failure states, and persistence.
4. Cover the critical path with automated tests.
5. Exercise a clean install from the README and fix every missing step.

## Done when

- A new user can go from clone to first successful workflow using only the README.
- The core workflow works without paid infrastructure unless the brief requires it.
- Tests cover the highest-risk behavior.
- Known limitations are explicit rather than hidden.

# Copy to .env and document every variable when it is introduced.
# Never put real credentials in this file.

APP_ENV=development
# Add only values required by the selected implementation.

# Dux-Soup product brief

## Problem

Mechanically this is a Playwright script that drives your own logged-in Chrome profile, clicks buttons, and waits a random number of seconds. An agent can get a working profile-visitor and connection-request sender running in a session, and a small campaign queue with daily caps by the end of a weekend. What you are actually buying from Dux-Soup is the accumulated caution: throttle curves that do not get your account restricted, selectors that keep working after LinkedIn reshuffles the DOM every few weeks, and a cloud mode so your laptop does not have to stay awake. Note that neither the paid tool nor your script is blessed by LinkedIn, so the ban risk is yours in both cases; you are just choosing who tunes the pacing. Build it if you want ten touches a day and enjoy fixing selectors, buy it if outreach volume is how you eat.

## Product outcome

Reads a CSV of LinkedIn profile URLs, drives your existing logged-in browser session to visit each one, optionally sends a connection request with a templated note, and logs outcomes to a local SQLite file with hard daily caps.

## Target user

A serious builder who needs a maintainable product foundation rather than a one-off demo.

## Required capabilities

- Node 20+ and a local Chrome or Chromium install
- A LinkedIn account you are willing to risk, logged in via a persistent browser profile
- A CSV of target profile URLs, sourced by you
- A machine that stays awake during campaign windows
- Tolerance for re-fixing selectors whenever LinkedIn ships a UI change

## Explicit non-goals for v1

- Pacing heuristics and safety limits tuned across a lot of accounts, not guessed at by you
- Selector maintenance: when LinkedIn changes markup, someone else patches their extension and nobody patches yours
- Cloud execution, so campaigns keep running with your laptop shut
- Webhook and CRM plumbing into HubSpot, Pipedrive and friends, plus Zapier glue
- Campaign reporting, reply detection and sequence branching that actually stops when someone answers

## Success criteria

- The primary workflow is measurable end to end.
- Setup is reproducible in a clean environment.
- Failure, recovery, and support paths are documented.
- Product claims match what the implementation actually guarantees.

# Architecture

## Starting brief

Build a local LinkedIn outreach assistant that drives my own browser. Stack: TypeScript, Node 20, Playwright with a persistent Chromium user data dir, better-sqlite3 for state, no server, no cloud, no telemetry, no account system. CLI only.

Structure:
- src/browser.ts: launches Playwright chromium with userDataDir from .env (PROFILE_DIR), headed by default so I can watch and take over.
- src/login.ts: opens linkedin.com and waits for me to log in manually, then exits. Never handle or store my password. No credentials in code or .env.
- src/db.ts: SQLite schema for targets (url, name, status, added_at) and actions (target_url, action_type, result, ran_at).
- src/import.ts: reads targets.csv (column: profile_url, optional first_name) and inserts new rows as status 'queued'.
- src/run.ts: the campaign loop.

Campaign loop rules:
- Read config from config.json: daily_visit_cap, daily_connect_cap, active_hours (start and end, local time), min_delay_seconds, max_delay_seconds.
- Default caps low and boring: 20 visits, 10 connects.
- Sleep a random interval between min and max delay between every action, plus a longer random pause every 5 actions.
- Stop immediately and log if the page shows a captcha, a checkpoint, an 'unusual activity' interstitial, or a login redirect. Do not retry.
- Actions: visit profile, and optionally click Connect and add a note rendered from templates/connect.txt with {{first_name}} substitution. Truncate notes to the field limit.
- Everything runs in dry-run mode unless --live is passed. Dry run logs the exact action it would take.
- Idempotent: never act twice on the same target for the same action type.

Also include: src/report.ts printing counts by status and action outcome, a README that states plainly that this automates a logged-in session and may violate LinkedIn's user agreement and can get an account restricted, and a .env.example with PROFILE_DIR only.

Out of scope: scraping or guessing email addresses, bulk profile data export, running headless to evade detection, proxy rotation, multiple accounts, reply detection, CRM sync, any hosted component.

## Boundaries

Separate the product into replaceable modules for interface, application logic, persistence, external integrations, and operational concerns. Keep domain logic independent from delivery frameworks and vendors.

## Production baseline

- Configuration: validated at startup with safe local defaults where possible.
- Security: least privilege, input validation, secret redaction, rate limits on abuse-prone paths, and no invented security primitives.
- Data: explicit schema and migrations, transactional writes where integrity matters, backup and restore instructions.
- Integrations: adapters around third-party providers, idempotent webhook or job processing, bounded retries, and timeouts.
- Observability: structured logs with request or operation IDs, an error-tracking hook, and health/readiness checks where a server exists.
- Quality: unit tests for domain rules, integration tests at module boundaries, and one end-to-end critical-path test.

## Decision records

For each major dependency, document why it was chosen, its failure mode, and how it can be replaced. Do not introduce infrastructure until a requirement justifies it.

# Agent instructions

- Read `PRODUCT.md` and `ARCHITECTURE.md` before changing code.
- Implement milestone by milestone; keep each change reviewable and leave the application runnable.
- Treat authentication, payments, encryption, imports, webhooks, and destructive actions as high-risk boundaries when present.
- Never invent cryptography or silently weaken a requirement to make a test pass.
- Use provider interfaces for external services and deterministic fakes in tests.
- Add migrations and rollback or recovery notes for persistent data changes.
- Log useful operational context without credentials, tokens, passwords, or personal data.
- Update documentation and run all checks before completing a milestone.

# Delivery milestones

## M0 — Decisions and scaffold

- Confirm the runtime, persistence model, threat boundaries, and deployment target.
- Create a reproducible local environment and continuous checks.

## M1 — Core workflow

- Implement the smallest end-to-end product path with validation and tests.
- Keep integrations behind interfaces.

## M2 — Trust layer

- Add secure failure behavior, recovery paths, audit-relevant events, and data safeguards.
- Test abuse cases and destructive operations.

## M3 — Operability

- Add structured logs, error reporting hooks, health signals, backup/restore documentation, and deployment configuration.

## M4 — Release gate

- Run a clean-install test, critical-path end-to-end test, dependency review, and documented rollback exercise.
- Compare the shipped behavior with `PRODUCT.md` and publish remaining limitations.

# Operations

## Before release

- Validate configuration and secrets at startup.
- Define backup, restore, and rollback procedures and test them.
- Document logs, error tracking, health signals, and alert ownership.
- Set dependency update and vulnerability review expectations.

## Incident checklist

1. Contain the issue without destroying evidence or user data.
2. Record the timeline and affected scope.
3. Rotate exposed secrets and revoke compromised sessions or credentials.
4. Restore from a verified source when needed.
5. Document the root cause, remediation, and regression test.

## Launch constraint

Do not market omitted Dux-Soup capabilities as implemented. The v1 non-goals in `PRODUCT.md` remain user-visible limitations until they are deliberately delivered.

$ choose a build depth, inspect the files, then open the complete pack in your agent · this prompt is generated from the build plan · improve it via PR