Skip to content
Worix
BrowsePublish
Log inSign Up

task-competitor-research

Multi-step TASK — researching competitors from public information — the ones the merchant names, or the three closest the server finds from their store when they ask you to find them — what they push, what they charge, what they are promoting and how they present themselves, each point with its source, and how they differ from the merchant. Enter when: the merchant names competitors to look at, research or compare, or pastes a competitor's link ("what is Gymshark promoting right now?", "look at alo.com"), or asks you to find their competitors — including their answer to the "name them or let me find them" question ("who are my competitors?", "Find them for me"). Do NOT enter: competitor research that names no competitor and does not ask you to find them ("Research my competitors") — ask first, as AGENTS.md says under Market & web research; general web research, trends or examples with no competitor to study (answer it directly with GOOGLE_SEARCH); how their own store, traffic, posts or ads are doing (that is a data question, or marketing advice); topic or content ideas — what to write or post about; a promotion plan or campaign ideas (that is task-promo-plan); greetings or small talk; setting up, auto-generating, calibrating, or re-scanning Brand IQ (that is task-brand-iq-setup). Read this skill's full body BEFORE your first action — it carries the node graph you must report progress against, one node at a time.

0 downloads
Free
Reviewed
internal
marketing
task

Generated from tasks/competitor-research/task.yaml by /task-authoring.

Do not edit here — edit the definition and regenerate.

name: task-competitor-research description: >- Multi-step TASK — researching competitors from public information — the ones the merchant names, or the three closest the server finds from their store when they ask you to find them — what they push, what they charge, what they are promoting and how they present themselves, each point with its source, and how they differ from the merchant. Enter when: the merchant names competitors to look at, research or compare, or pastes a competitor's link ("what is Gymshark promoting right now?", "look at alo.com"), or asks you to find their competitors — including their answer to the "name them or let me find them" question ("who are my competitors?", "Find them for me"). Do NOT enter: competitor research that names no competitor and does not ask you to find them ("Research my competitors") — ask first, as AGENTS.md says under Market & web research; general web research, trends or examples with no competitor to study (answer it directly with GOOGLE_SEARCH); how their own store, traffic, posts or ads are doing (that is a data question, or marketing advice); topic or content ideas — what to write or post about; a promotion plan or campaign ideas (that is task-promo-plan); greetings or small talk; setting up, auto-generating, calibrating, or re-scanning Brand IQ (that is task-brand-iq-setup). Read this skill's full body BEFORE your first action — it carries the node graph you must report progress against, one node at a time. kind: task task_name: "Competitor research" label_template: "<competitors> research" tags: [internal, marketing, task] license: MIT-0 nodes:

  • id: pick label: "Finding stores related to your catalogue" purpose: "Start the research with FIND_COMPETITORS — passing only the competitors the merchant named, or none so the server picks three from their store — and run every search it returns with GOOGLE_SEARCH. Who was settled before the task, by what the merchant named or asked you to find; nobody is asked how deep." advance_by: agent tools:
    • yepairag_marketing__FIND_COMPETITORS
    • yepairag_marketing__GOOGLE_SEARCH next:
    • to: research when: "FIND_COMPETITORS has started the research and every search it returned has run (or it returned none)"
  • id: research label: "Reading their public websites" purpose: "Call RUN_COMPETITOR_RESEARCH once: the server settles who, reads each one's public website and the merchant's own store. Then search the web, at most twice each, only for the competitors its next names." advance_by: agent tools:
    • yepairag_marketing__RUN_COMPETITOR_RESEARCH
    • yepairag_marketing__GOOGLE_SEARCH next:
    • to: report when: "the facts are in — the research has at least one competitor read or searched (its data_state has_data is true)"
    • to: pick when: "before the report is up, the merchant names other competitors to research"
  • id: report label: "Comparing prices and promotions with yours" purpose: "Put the report card up: per competitor how they present themselves, what they push and what they promote now, and two or three moves to borrow — every sentence citing the facts it stands on; the card adds how they differ from the merchant by itself. The card is the reply and ends the task." advance_by: agent artifact: { type: competitor_report } tools:
    • yepairag_marketing__SHOW_COMPETITOR_REPORT

Task — Competitor research

Flow

flowchart TD
  pick["Finding stores related to your catalogue"] --> research["Reading their public websites"]
  research --> report["Comparing prices and promotions with yours"]
  research --> pick
  classDef terminal stroke-width:3px
  class report terminal

Rounded = waiting on the merchant · square = you are working · bold border = terminal.

pick

Finding stores related to your catalogue — Start the research with FIND_COMPETITORS — passing only the competitors the merchant named, or none so the server picks three from their store — and run every search it returns with GOOGLE_SEARCH. Who was settled before the task, by what the merchant named or asked you to find; nobody is asked how deep.

  • FIND_COMPETITORS starts every research; who was settled before the task, and nobody is asked how deep. Pass in named only the competitors they named — in this request or in their answer to the "name them or let me find them" question — as they typed them, names or website addresses, and leave it out when they asked you to find them: the server then picks the three closest from their store. Never suggest or add a competitor yourself, and never top a named list up to three. When they turn down the server's picks and name others ("try Veja instead"), the ones they named are the whole research.
  • Run every query in its searches with GOOGLE_SEARCH, as written — one call each, nothing else — then go to research. Pass nothing on from what the searches found: the server reads what they cited by itself. No searches means straight to research. NO_STORE_TO_MATCH means there is no store to match against: ask in one sentence which brands they compete with, and start again with the ones they name.

research

Reading their public websites — Call RUN_COMPETITOR_RESEARCH once: the server settles who, reads each one's public website and the merchant's own store. Then search the web, at most twice each, only for the competitors its next names.

  • RUN_COMPETITOR_RESEARCH once, with no arguments — the server settles who (each pick it made carries why_picked) and reads their public sites and the merchant's own store. Then call GOOGLE_SEARCH only for the competitors its next names — the ones whose site could not be read or found — at most twice each: what they sell and at what prices; what they are promoting now and how they present themselves. A competitor whose site was read is not searched.
  • A site that could not be read is reported as it is. Its note says why; never guess what it would have said and never fill its gap from memory. When nobody was found (NO_COMPETITORS_FOUND, data_state.has_data false), one sentence says so and asks which brands they see as their competitors — that is "never stop at I can't".

report

Comparing prices and promotions with yours — Put the report card up: per competitor how they present themselves, what they push and what they promote now, and two or three moves to borrow — every sentence citing the facts it stands on; the card adds how they differ from the merchant by itself. The card is the reply and ends the task.

  • One SHOW_COMPETITOR_REPORT call is the whole reply — no text before or after it. For each competitor, what you can support: positioning (how they present themselves and to whom), pushing (what they lead with), promotions (what they run now). Then two or three moves the merchant could borrow, each a short instruction with a one-sentence why that says what the cited fact shows and offers the payoff as something to try, never a promised result. summary is your one-sentence key takeaway. How they differ from the merchant is the card's own table, worked out by the server — not a sentence of yours.
  • Every sentence cites the keys it stands on — D… and O… from the research, S… refs from GOOGLE_SEARCH. A figure in a sentence appears exactly as the fact it cites states it; never a total, percentage or comparison of your own (the research already worked the comparisons out).
  • A competitor's traffic, sales, revenue and ad spend are never stated or estimated — they are not public, so any number would be invented. The Ad Library link on the card is the way to see their ads.
  • A refusal lists every problem: fix all of them and call again, saying nothing.

How a task runs

This part is the same for every task; only the nodes differ.

1 — Entering the task. Report the first node before doing anything else. That report is what opens the run: it tells the backend a task has started and gives the merchant their first progress line.

2 — Entering each node. Report it the moment you enter, before doing that node's work. One report per node. A single turn usually passes through several nodes with no input from the merchant in between — reporting as you go is what lets them watch the work happen instead of staring at nothing until you finish. A node whose work is a single tool call still gets its report first.

3 — Reaching the end. A terminal node is one with no way out except back to itself. Reporting it completes the task — the backend works that out from the graph, so there is nothing extra for you to send.

4 — After completion. A completed task is not sealed. If the merchant later asks to change something, report the node that owns it again and the task is simply unfinished again. Never open a second task for a change to work you already did.

How to report

yepairag_task__REPORT_TASK_STEP
  task: "competitor-research"
  node: "<the node you just entered>"
  • Two fields, both read straight off the graph above. The step's label, its number, the total and whether the task just completed are filled in by the backend — you cannot get them wrong because you do not send them.
  • The reply carries legal_next: the only nodes you may report next.
  • ok: false means nothing was recorded and the run has not moved. Correct the node id and report again rather than carrying on as if it had been accepted.
  • Never restate any of it to the merchant. It drives a progress bar they are already looking at.

Download

ZIP package — ready to use

Skill Info

Creator
Yepai
License
MIT-0
Downloads
0
Published
Nov 20, 2026
Updated
Nov 20, 2026