lennoxsaint/Threadify-workflows / plugins/claude/skills/threadify-inbound-replies
threadify-inbound-replies
VerifiedConnect Threadify MCP, review a backlog of Threads comments and mentions grouped beneath full posts, edit five cards or all comments, explicitly send approved replies, and learn from verified edits. Supports interrupted batches and custom periods.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.8 KB |
| references/inbound-contract.md | markdown | 6.5 KB |
| references/threadify-001.md | markdown | 3.7 KB |
| references/workflow-manifest.json | config | 3.1 KB |
| references/workflow-readme.md | markdown | 11.5 KB |
| scripts/app.js | script | 8.6 KB |
| scripts/cli.mjs | script | 1.9 KB |
| scripts/index.html | other | 1.6 KB |
| scripts/server.mjs | script | 2.9 KB |
| scripts/store.mjs | script | 23.2 KB |
| scripts/style.css | other | 3.3 KB |
references/inbound-contract.md
View on GitHub# Local packet interface v1
Run from the repository root:
```sh
node tools/inbound/cli.mjs configure --packet /path/to/config.json
node tools/inbound/cli.mjs window --account @example
node tools/inbound/cli.mjs import --packet /path/to/snapshot.json
node tools/inbound/cli.mjs drafts --packet /path/to/drafts.json
node tools/inbound/cli.mjs serve --account @example
node tools/inbound/cli.mjs status --account @example
```
All mutations take `--packet FILE`; packets contain `version: 1` and `account`. All commands support `--root DIRECTORY` for isolated state. Default state is in the user's `.local/share/threadify-workflows/inbound` directory, outside the checkout. Output is JSON, errors exit nonzero. Status includes private bodies; do not redirect it into public logs. Root directory is private (0700), files are 0600. Atomic rename/fsync and exclusive locks protect transactions; errors leave the previous file intact. An interrupted lock fails closed. Inspect its PID, confirm the process is dead, then remove only that account's stale lock before retrying. Never delete a live writer's lock.
## Configure and discovery
`configure`: `{version:1, account:"@example", preference:"five"|"all", timezone:"Europe/London"}`. Ask for the review preference before configuring. `window` supports `--period 24h|7d`, `--start ISO`, `--end ISO`; default is since checkpoint, first use 24h. Resume an open batch first.
`import` takes:
```json
{
"version": 1,
"account": "@example",
"start": "2026-01-01T00:00:00Z",
"end": "2026-01-02T00:00:00Z",
"coverage": {"complete": true, "gaps": []},
"exclude_ids": [],
"items": [{
"source_id": "synthetic-comment-1",
"lane": "comment",
"author": "@reader",
"text": "Full verbatim comment",
"url": "https://www.threads.com/example-comment",
"occurred_at": "2026-01-01T09:00:00Z",
"pending": true,
"available": true,
"safety": "safe",
"context_complete": true,
"ancestry": [],
"post": {
"id": "synthetic-post-1",
"verified": true,
"text": "Complete post or ordered thread",
"url": "https://www.threads.com/example-post",
"metrics": {"views": 12, "likes": 2},
"metrics_checked_at": "2026-01-02T00:00:00Z",
"metrics_as_of": null
}
}]
}
```
Examples are synthetic. `source_id` must be native in real runs. Ancestry is an array of `{author,text}`. Unknown URLs/metrics are null or omitted. Unsafe sources need identity/timestamp/eligibility only; omit their bodies. Context completeness includes relevant media. The output gives stable item `id` and `revision`; use those values unchanged. Frozen batch reimports are idempotent and do not refresh the cohort.
## Enrich or refresh a captured item
`context`: `{id,revision,evidence_ref,source:{source_id,author,safe,pending,available,text,post,ancestry,context_complete,url}}`. This updates the same native target after live or browser context enrichment; it does not reopen or expand the frozen cohort. An established parent ID and author cannot change; a previously unknown parent may be bound by verified enrichment. Refreshing revokes approval. Ineligible sources become unavailable. In-flight sends must be reconciled instead of refreshed.
## Draft and edit
`drafts`: add `drafts:[{id,revision,source_id,text,mode:"native"|"agent",context_verified:true,gates_passed:true}]`. This retains the immutable first draft plus draft/edit history and the latest baseline. Regeneration requires current revisions; never overwrite unsaved user edits.
`edit`: add `{id,revision,text}`. Text is stored without trimming or normalization. Changed text increments revision and revokes approval. `decide`: add `{id,revision,action:"approve"|"defer"|"skip"|"undo"|"regenerate"}`. Empty approvals are rejected. A decision increments revision. UI actions and agent commands use the same store.
`surface`: add `capabilities:{create:true,stableIdentity:true,readback:true}` only when verified. All three true selects native; otherwise local. `serve` starts a loopback-only server and returns a private session URL. It has no send endpoint. A stale edit gets a conflict rather than overwriting the newer revision. Copy unsaved text before refreshing after a conflict.
## Prepare/send/readback
`prepare-send`: add `authorization:{explicit:true,reference:"user-message-reference"}` and `items:[{id,revision,source_id,author,final_hash,pending:true,safe:true,available:true,gates_passed:true,evidence_ref,checked_at}]`. Compute `final_hash` with SHA-256 of exact UTF-8 text (exported `hash` helper). `quote_target_kind` may be `comment` or `mention` only when the provider verifies that mapping. Proof fields describe actual reads, not invented assertions. This command must follow the user's chat send instruction; local approval alone is insufficient.
Output contains the durable `attempt.key` and exact prepared reply payload. The agent must check that `send_replies` or `send_reply` is listed in the current session and call it with that payload; where the connection requires an approval, it comes from `review_reply` first. `record-send`: `{key,receipt,eta}`. Receipts are stable provider evidence references, not local success labels.
`reconcile`: `{key,results:[{id,source_id,author,status,evidence_ref,checked_at,...}]}`. Status is:
- `published`: additionally requires exact `text`, `reply_id`, `url`, `pending_before:true`, `replied_after:true` and a recorded send receipt.
- `failed`: requires `definitely_not_sent:true` and `failure_receipt`; returns the item to unapproved drafted state for a fresh check and approval.
- `uncertain`: blocks another send and retains the item for further readback.
The `feedback` entry `{key,receipt}` records a successful provider feedback call. Pending feedback payloads are in status and remain retryable independently of publication. `close`: optional `batch_id`; reports coverage/checkpoint and feedback backlog. `status --batch ID` reads a historical batch's current item states.
The helper trusts the authorized agent to gather actual MCP evidence. It checks identity, revisions, hashes, timestamps and required proof fields; it cannot independently authenticate provider receipts or grant sending authority.
`receipt --account @example` emits a body-free review/send receipt: hashes, revisions, proof references, coverage and remaining feedback. The operator adds actual editor surface evidence; the helper explicitly reports that field as requiring readback. Provider references and account handles remain private operational metadata unless redacted for a public example.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |