Aidelly

Build me a personal social publishing agent: an MCP server (TypeScript, official SDK) exposing tools post_now, schedule_post, list_queue and cancel_post. Posts go to Bluesky (atproto, app-password auth, image + video upload via the video service) and Mastodon (masto.js, app registration). schedule_post writes to a SQLite queue; a cron worker publishes due posts with retry and exponential backoff, and failed posts surface through a get_failures tool instead of silently dropping. Include a dry-run mode that previews the per-platform rendering (300-grapheme limit handling for Bluesky, 500 chars for Mastodon) before anything goes live.

# AGENTS.md — Build guide for Aidelly

## Project scope
An MCP server exposing post-now and schedule tools that publish text and images to Bluesky and Mastodon from a cron-backed queue.

Catalogue verdict: no. The agent layer is the buildable part: a weekend of vibe coding gets you an MCP server that posts to Bluesky and Mastodon, whose APIs are open. Everything after that is the product: approved OAuth apps for Meta, TikTok, YouTube, LinkedIn, X and Google Business (app review processes measured in weeks, and X's API is pay-per-use), token refresh across eleven platforms that silently break in eleven different ways, publish retry/recovery infrastructure, per-client workspaces with approval workflows, and white-label so agency clients never see the tool. The integration upkeep is not a one-shot; it is the job.
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- Do not publish, spend money, contact customers, delete source data or run irreversible migrations without the project owner's authorization.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
- Bluesky app password + Mastodon app registration
- hosted scheduler/cron
- database for queue and post state
- media storage
- developer app approval from each closed platform (Meta, TikTok, YouTube, X) for anything beyond the open ones

## Stack and architecture
- Build me a personal social publishing agent: an MCP server (TypeScript, official SDK) exposing tools post_now, schedule_post, list_queue and cancel_post. Posts go to Bluesky (atproto, app-password auth, image + video upload via the video service) and Mastodon (masto.js, app registration). schedule_post writes to a SQLite queue; a cron worker publishes due posts with retry and exponential backoff, and failed posts surface through a get_failures tool instead of silently dropping. Include a dry-run mode that previews the per-platform rendering (300-grapheme limit handling for Bluesky, 500 chars for Mastodon) before anything goes live.
- Model sources or channels, posts, dated observations, drafts, and reports; keep stable source IDs and timestamps.
- Interface for this agency social media via AI agents workflow: a focused agency social media via AI agents input, review, and export interface.

## Security and data integrity
- On invalid input or interrupted processing, keep the original record, show the failed step, and permit a safe retry.
- Validate imported records and file paths; keep user data within the storage location named in the prompt.
- State the practical limit: the closed platforms: Meta, TikTok, YouTube, X and Google Business need approved developer apps and policy compliance.

## Agent implementation rules
- Project rule, data: Model sources or channels, posts, dated observations, drafts, and reports; keep stable source IDs and timestamps.
- Project rule, behavior: An MCP server exposing post-now and schedule tools that publish text and images to Bluesky and Mastodon from a cron-backed queue.
- Project rule, recovery: On invalid input or interrupted processing, keep the original record, show the failed step, and permit a safe retry.

## 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 agency social media via AI agents operation with sample or user-owned data: An MCP server exposing post-now and schedule tools that publish text and images to Bluesky and Mastodon from a cron-backed queue.
- Show a concrete result or export: AI agent that runs client social media accounts: publish, approvals, DMs, reports, driven from Claude/ChatGPT via MCP or its own app.
- State the practical limit: the closed platforms: Meta, TikTok, YouTube, X and Google Business need approved developer apps and policy compliance.

## Engineering roadmap
1. Phase 1, architecture and data: Build me a personal social publishing agent: an MCP server (TypeScript, official SDK) exposing tools post_now, schedule_post, list_queue and cancel_post. Posts go to Bluesky (atproto, app-password auth, image + video upload via the video service) and Mastodon (masto.js, app registration). schedule_post writes to a SQLite queue; a cron worker publishes due posts with retry and exponential backoff, and failed posts surface through a get_failures tool instead of silently dropping. Include a dry-run mode that previews the per-platform rendering (300-grapheme limit handling for Bluesky, 500 chars for Mastodon) before anything goes live. Model sources or channels, posts, dated observations, drafts, and reports; keep stable source IDs and timestamps.
2. Phase 2, implement: An MCP server exposing post-now and schedule tools that publish text and images to Bluesky and Mastodon from a cron-backed queue.
3. Phase 3, implement: AI agent that runs client social media accounts: publish, approvals, DMs, reports, driven from Claude/ChatGPT via MCP or its own app.
4. Phase 4, review and output: An MCP server exposing post-now and schedule tools that publish text and images to Bluesky and Mastodon from a cron-backed queue. AI agent that runs client social media accounts: publish, approvals, DMs, reports, driven from Claude/ChatGPT via MCP or its own app.
5. Phase 5, recovery and acceptance: On invalid input or interrupted processing, keep the original record, show the failed step, and permit a safe retry. Verify this invariant with a saved fixture: An API gap must be labelled missing; a replayed export cannot double-count the same observation. State the practical limit: the closed platforms: Meta, TikTok, YouTube, X and Google Business need approved developer apps and policy compliance.

## Paid-product capabilities outside this build
- the closed platforms: Meta, TikTok, YouTube, X and Google Business need approved developer apps and policy compliance
- client approval workflows and per-client workspace isolation
- white-label (your clients seeing your brand, not your stack)
- publish reliability infrastructure: retries, token-refresh, failure emails, watchdogs
- inbox/DM automation and cross-client analytics reports

## Implementation prompt
Build me a personal social publishing agent: an MCP server (TypeScript, official SDK) exposing tools post_now, schedule_post, list_queue and cancel_post. Posts go to Bluesky (atproto, app-password auth, image + video upload via the video service) and Mastodon (masto.js, app registration). schedule_post writes to a SQLite queue; a cron worker publishes due posts with retry and exponential backoff, and failed posts surface through a get_failures tool instead of silently dropping. Include a dry-run mode that previews the per-platform rendering (300-grapheme limit handling for Bluesky, 500 chars for Mastodon) before anything goes live.

## 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.