A missing skill and an unfashionable label are not the same gap
Annual curriculum reviews compare a syllabus with the requirements that employers state. The result is usually a list of mostly irrelevant items. A tool mentioned in advertisements this year can be a rebranded version of one the program already teaches. A true competency gap might never appear as a keyword in job titles.
To tell the two apart, you read the advertisement to find out what the work actually entails, then look at the curriculum for the outcomes, not the module names. A module named Database Systems might already include what the advertisement calls data modeling.
A second judgment is involved. A gap in six percent of the advertisements in a market where you do not place yourself matters less than a gap in one third of the advertisements from your three main employers.
The agent creates a shortlist of well-reasoned recommendations for the curriculum committee. The agent makes no changes and gives no approvals. The decision on whether something should be taught stays where it belongs.
How the agent is built in Agent Studio
The prompt specifies the occupations graduates enter, including the boards, employers, and geographic areas, so the agent can refer to the actual destinations of your graduates rather than a global average.
Prompt skills also include your program vocabulary, which links outcome statements to the terms that advertisements use, so a rename does not count as a gap.
The system prompt states that each suggested gap must refer to the advertisements it is based on and name which part of the curriculum was compared with it, even when the match is partial.
The system prompt does not allow sorting based only on frequency. A gap must be justified by employer relevance and the consequences of the placement, or it is treated as an observation rather than a recommendation.
A scheduled flow gathers the advertisements through the Scrappey and HTTP Request nodes, then passes them to the Agent Chat node along with the curriculum read from Notion.
Google Sheets receives the analysis. Each row contains a proposed gap and supporting evidence, in a format that a curriculum committee can work through.
The Slack node provides the shortlist to the program lead. The agent does not edit any curriculum document.
What this agent is built from
- Prompt skills: declared job boards, employer set, and geography for your graduates.
- Prompt skills: program vocabulary mapping outcome statements to market terms.
- System prompt: every gap cites its adverts and names what was checked in the curriculum.
- Scrappey (Util) with HTTP Request (Util): advert collection from the declared sources.
- Agent Chat (Util): receives adverts together with the curriculum read from Notion.
- Google Sheets (Integration): one row per proposed gap, with its evidence.
- When an ActionFlow is enough: If counting which declared skills appear in adverts and which do not is the work, build the workflow. That count is a reasonable annual baseline.
Frequently asked questions
Start building AI workflows
Create a free account, open a template or a blank canvas, and run your first ActionFlow.
Newsletter
Get product updates
New nodes, agents, and product notes. We send mail only when we have something worth opening.