SubSignal

Build me a focused Reddit monitoring for buying intent workflow for the personal core of SubSignal. Requirements:

- Use exactly this stack: Node.js 22 + TypeScript + SQLite. Store every item you have seen in SQLite so the same thread is never delivered twice, including across restarts.
- Paid product context: Watches Reddit for people describing the problem you solve, and surfaces them as leads. Build only this DIY scope: Poll a list of subreddits on a schedule, filter new posts by keyword, ask one model whether each is a buying signal for your product, and send the hits to Slack or email.
- For Reddit monitoring for buying intent, import only consented contacts with source, suppression status, and a stable external ID.
- Draft a sequence, preview personalization, and queue a send through an authorized provider.
- Track provider receipts, replies, and opt-outs beside the original contact.
- Use a focused Reddit monitoring for buying intent input, review, and export interface. Required input or access: Reddit API credentials, within Reddit's terms of use; OpenAI or Anthropic API key in .env. Keep credentials in .env.
- Acceptance: with one labelled sample, show the input, saved intermediate state, and exported result; verify this invariant: An opt-out suppresses every future step; retrying a timed-out send cannot silently duplicate it.
- Out of scope: a historical archive you cannot backfill from the live API; ranking a lead against everything seen before, instead of judging it alone. Keep this a personal, inspectable workflow.
- Include a README with setup, a sample input, required keys or permissions, data location, and the supported scope.

# AGENTS.md — Build guide for SubSignal

## Project scope
Poll a list of subreddits on a schedule, filter new posts by keyword, ask one model whether each is a buying signal for your product, and send the hits to Slack or email.

Catalogue verdict: no. Do not mistake the loop for the product. Polling subreddits and asking a model "is this a buying signal?" is an evening, and it is also the part that does not matter. What makes this useful is the accumulated archive behind it: years of threads at a scale you cannot backfill, which is what lets it tell a recurring complaint from a one-off and rank a lead against everything already seen. Your build starts empty on day one and stays empty, because Reddit will not sell you the past. That is a data problem, not a code problem.
Use the implementation prompt below to define the deliverable. Complete each phase's acceptance checks before extending the scope.

## Working agreement
- Inspect the repository and its existing instructions before choosing paths, dependencies or commands. Keep one coherent stack and explain changes to the proposed architecture.
- Plan a vertical slice that accepts a real input and produces the useful output described below. Persist only the state the prompt calls for; respect memory-only and upstream-managed workflows. Use fixtures only when they are clearly labelled.
- After scaffolding, document the actual install, development, check and build commands in README and keep them synchronized with the package or project manifest. Do not report commands as successful unless they ran.
- Work in small steps. At handoff, list implemented flows, checks actually performed, remaining blockers, and any credentials or provider setup the owner must supply.
- Do not publish, spend money, contact customers, delete source data or run irreversible migrations without the project owner's authorization.

## Prerequisites
- Reddit API credentials, within Reddit's terms of use
- OpenAI or Anthropic API key in .env
- Node.js 22 and a scheduler such as cron
- SQLite to remember what you have already seen

## Stack and architecture
- Use exactly this stack: Node.js 22 + TypeScript + SQLite.
- Store every item you have seen in SQLite so the same thread is never delivered twice, including across restarts.
- Interface for this Reddit monitoring for buying intent workflow: a focused Reddit monitoring for buying intent input, review, and export interface.

## Security and data integrity
- An opt-out suppresses every future step; retrying a timed-out send cannot silently duplicate it.
- Keep API keys and OAuth secrets in .env; redact them from logs, generated files, and error responses.
- State the practical limit: a historical archive you cannot backfill from the live API.

## Agent implementation rules
- Project rule, data: Store every item you have seen in SQLite so the same thread is never delivered twice, including across restarts.
- Project rule, behavior: For Reddit monitoring for buying intent, import only consented contacts with source, suppression status, and a stable external ID.
- Project rule, recovery: An opt-out suppresses every future step; retrying a timed-out send cannot silently duplicate it.

## Optional agent skills and references
No installable skill is listed for this guide; use the project rules and the stack documentation.

Read the linked SKILL.md and its dependencies before adding a skill. Select only the skills matching this project's runtime and task; their documentation does not supply API access, credentials or approval to perform external actions. Pin the reviewed revision where the tool supports it. Follow the chosen agent's documented project-level installation mechanism.

## Distribution ideas
These are optional planning notes. Obtain the owner's approval before publishing or contacting anyone.
- Demonstrate this Reddit monitoring for buying intent operation with sample or user-owned data: For Reddit monitoring for buying intent, import only consented contacts with source, suppression status, and a stable external ID.
- Show a concrete result or export: Track provider receipts, replies, and opt-outs beside the original contact.
- State the practical limit: a historical archive you cannot backfill from the live API.

## Engineering roadmap
1. Phase 1, architecture and data: Use exactly this stack: Node.js 22 + TypeScript + SQLite. Store every item you have seen in SQLite so the same thread is never delivered twice, including across restarts.
2. Phase 2, implement: For Reddit monitoring for buying intent, import only consented contacts with source, suppression status, and a stable external ID.
3. Phase 3, implement: Draft a sequence, preview personalization, and queue a send through an authorized provider.
4. Phase 4, review and output: Track provider receipts, replies, and opt-outs beside the original contact.
5. Phase 5, recovery and acceptance: Verify this invariant with a saved fixture: An opt-out suppresses every future step; retrying a timed-out send cannot silently duplicate it. State the practical limit: a historical archive you cannot backfill from the live API.

## Paid-product capabilities outside this build
- a historical archive you cannot backfill from the live API
- ranking a lead against everything seen before, instead of judging it alone
- a curated subreddit and keyword set you did not have to discover
- deduplication and history across runs
- someone tracking Reddit API terms and rate-limit changes for you

## Implementation prompt
Build me a focused Reddit monitoring for buying intent workflow for the personal core of SubSignal. Requirements:

- Use exactly this stack: Node.js 22 + TypeScript + SQLite. Store every item you have seen in SQLite so the same thread is never delivered twice, including across restarts.
- Paid product context: Watches Reddit for people describing the problem you solve, and surfaces them as leads. Build only this DIY scope: Poll a list of subreddits on a schedule, filter new posts by keyword, ask one model whether each is a buying signal for your product, and send the hits to Slack or email.
- For Reddit monitoring for buying intent, import only consented contacts with source, suppression status, and a stable external ID.
- Draft a sequence, preview personalization, and queue a send through an authorized provider.
- Track provider receipts, replies, and opt-outs beside the original contact.
- Use a focused Reddit monitoring for buying intent input, review, and export interface. Required input or access: Reddit API credentials, within Reddit's terms of use; OpenAI or Anthropic API key in .env. Keep credentials in .env.
- Acceptance: with one labelled sample, show the input, saved intermediate state, and exported result; verify this invariant: An opt-out suppresses every future step; retrying a timed-out send cannot silently duplicate it.
- Out of scope: a historical archive you cannot backfill from the live API; ranking a lead against everything seen before, instead of judging it alone. Keep this a personal, inspectable workflow.
- Include a README with setup, a sample input, required keys or permissions, data location, and the supported scope.

## Completion evidence
Demonstrate the prompt's acceptance scenarios against the scoped workflow. Include setup from a clean checkout and failure recovery. Check persistence across restart and export/restore only for the state the prompt says to store; for memory-only tools, confirm that temporary content is discarded as specified. Record actual results and remaining limitations. A detailed plan alone does not establish a working replacement.