research-then-write
Skill0 starsChanged 6 days agoMITView on GitHub
Find posts that are working in the user's niche on Instagram, TikTok or YouTube, or look at a creator's recent posts, save the best ones as inspiration, then remake one as an original piece in the user's voice that borrows the structure, not the words. Use when the user asks what is going viral, wants ideas from a competitor or creator, or wants to remake a successful post their own way. Research and writing use credits, so it confirms before each paid step.
In this repository3 other files it gives its agents.
- setupSkill
- weekly-content-planSkill
- write-in-my-voiceSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 36
If a search returns a job_id, it is still running: read it later with get_viral_results or get_competitor_results.
- Distinctive: line 52
Write a short brief in your own words: the structure to borrow, the user's own topic and angle, and what to leave behind.
- Very specific: line 54
Call estimate_cost for the formats the user wants.
- Very specific: line 56
Call generate_content with remix pointing at the post: { "saved_key": ...
- Very specific: line 61
If you get a job_id, use get_generation until it is done.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Research, then write
Research what works, then write something original in the user's voice. Borrow the structure (hook type, pacing, format, angle), never the wording, jokes or personal stories of the original.
1. Choose the research path
- Viral posts in the niche:
find_viral_postswith aplatform(instagram,tiktokoryoutube). Leavenicheout to use the niche from the user's profile, or set it, pluswho,keywords(up to 5),window(such as30d) andlanguage, when the user is specific.load_more_viralwith the samejob_idgets more results. - What tracked creators posted lately:
get_feed(free), best first, optionally for oneplatform. - One tracked creator in depth:
list_tracked_creators, thenget_creator_reportwith thecreator_idfor their content pillars and top posts (add apost_idfrom the report for one post's full script and analysis). Both are free. - New creators to track:
track_creatorwith onecreator(an @handle or profile link, plusplatformfor a bare handle) or up to 5 at once increators.untrack_creatorfrees a slot. - Creators to watch:
find_competitorswith aplatform(plusfocus, such as "real estate agents in Dubai", andlanguagewhen the user is specific), thenget_competitor_resultswith thejob_idif it is still running. - Posts saved earlier:
list_inspiration(free), optionally bystatus:saved(To remake),usedorarchived.
find_viral_posts, load_more_viral, track_creator (a new scan costs credits per creator; one
scanned recently costs nothing), find_competitors and get_transcript (when the transcript is not
stored yet) use credits. Their prices are in get_account. Before any of them, tell the user the
cost and ask. Reading results (get_viral_results, get_competitor_results, get_creator_report,
list_tracked_creators, get_feed, list_inspiration) is free.
If a search returns a job_id, it is still running: read it later with get_viral_results or
get_competitor_results. A first creator scan takes a few minutes.
2. Pick what to learn from
Show the user the three to five most relevant results: a link, the hook, and in one line what made it work. Ask which one to build on.
- To keep a post for later,
save_inspiration(free) withjob_idandpost_key(a viral search result) orpost_id(a tracked creator's post), and anotesuch as "copy the cold open". It returns asaved_key.unsave_inspirationremoves one. - For a video,
get_transcriptgives the spoken words when the structure matters: setjob_idandpost_key, or asaved_key. Confirm the cost first if it is not free.
3. Remake it as an original piece
- Write a short brief in your own words: the structure to borrow, the user's own topic and angle, and what to leave behind.
- Call
estimate_costfor the formats the user wants. Tell them the cost and ask before writing. Do not callgenerate_contentuntil the user confirms. - Call
generate_contentwithremixpointing at the post:{ "saved_key": ... },{ "job_id": ..., "post_key": ... }or{ "post_id": ... }. A remix keeps that post's structure, writes new words in the user's voice and has no link fee. Put the user's angle and what to leave behind ininstructions. For a post found elsewhere, send its link assource_urlinstead, or send the brief asidea. - If you get a
job_id, useget_generationuntil it is done. - Once the remake is written,
save_inspirationwith thesaved_keyandstatus: "used"moves the original off the To remake list.
4. Present and save
Present the draft as the write-in-my-voice skill does (summarize when the app shows a drafts card),
with one line on which post inspired it and what was borrowed. Remind the user to make it their own
before posting. Offer to save it with save_to_library (free) with the draft_id and the hook they
picked.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

