Skip to content

lennoxsaint/Threadify-workflows / plugins/claude/skills/threadify-run-my-threads

threadify-run-my-threads

Verified

Set up one daily schedule so your agent prepares five Threadify-written Threads posts every morning from a weekly plan (one long-form thread on six days, one greatest-hit repost a day, hook-first structures and a chosen CTA level), shows them in this thread, and schedules them only after you reply "yes". Use when someone asks Threadify to run their Threads, prepare posts every morning for approval, or says "I want to approve posts, not write them".

references/state-and-recovery.md

View on GitHub
# Private state and recovery contract

Use `node scripts/run-state.mjs /absolute/private/run.json` with one JSON event on stdin. It atomically persists each event under an exclusive lock (a lock error means retry after reading state; never delete a live lock). Files are owner-only. Do not put private state or source bodies in Git. Existing historical cards remain unchanged; use this format for new runs.

Events: `init` (account, expires, slots with id/time), `assign` (slot, assignment), `hold` (slot, reason), `begin` (slot, unique marker, at), `result` (slot, draft_id, parts, optional plug/model), `failure` (slot, error/category), `recover` (slot, complete, drafts), `edit` (slot, instruction, edits), `casing` (slot, choice, keep), `repost` (slot, source_id, eligible, parts), `approve` (hash, at), `schedule` (slot, at, account, calendar_complete, occupied), `readback` (slot, account, time, parts, plug, id, list_verified). Include expected_hash for concurrent text edits.

Assignments contain source {url, text, performance, observed_at}, template, adaptation, angle, mechanism, proof. Evidence identifiers must name the same underlying proof consistently; changing a label does not make repeated evidence distinct. The helper checks structure and exact duplicates, not truth or semantic novelty. Read and assess sources before generation, not in an output audit loop.

Persist begin before the call; include its unique marker in generation input. Persist result immediately. On timeout use list_drafts then get_draft, fully paginating the relevant interval. Match the exact attempt marker, account and creation time, never a guessed title. If the provider does not preserve a usable correlation marker, the search is incomplete: hold rather than risk duplication. One confident match recovers success. A complete search with no match permits the configured fallback. Multiple matches or inaccessible recovery holds the slot. Never retry a saved slot. Stop on usage/rate/input/account failures. Report actual returned model, otherwise “not reported”.

Edits use part index, start/end offsets, expected exact selected text and replacement. Hook-only edits also use scope hooks and hook_end from the owner's selected hook. Replay on the latest revision only. Preserve a before/after snapshot and owner instruction. Deletion is replacement with an empty string. No inferred extra cleanup. Casing is applied separately using casing-guard against the latest wording revision, with originals/finals and guard evidence retained privately; do not treat a casing guard as proof that wording edits were authorized. Verbatim greatest hits remain unchanged.

Approval is the full revision hash bound to account, all parts, plug and times. Show all five slots, including held reasons, full text and local/UTC times. A requested edit clears approval. Use exact text/posts without draft_id after edits or casing, because a stored draft can contain older copy. Persist schedule pending before calling the provider with the stable key. Reconcile pending outcomes before retrying, never replay a whole batch. Platform validation remains required. Invalid text is held unchanged.

Read back get_schedule_status and list_scheduled_posts. Exact mismatches stop, never silently pass. For provider-transformed CTA URLs, retain raw response and explain the transformation; do not declare an exact match or modify account tracking. A missing Auto Plug echo is also a mismatch. Cancellation needs a separate owner request. Existing schedules are never changed by this workflow.
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: 471kUpdated Claude CodeCodexVerifiedInstall
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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
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 availableStars: 322kUpdated Claude CodeCodexVerifiedInstall