CtrlK
BlogDocsLog inGet started
Tessl Logo

jbaruch/auto-skill-discovery

Automated pipeline that takes a company name and produces a custom Tessl skill plus an eval report showing per-scenario lift (baseline agent vs with-skill agent). A1 MVP cell of the produce/consume × personalization 2x2.

88

1.45x
Quality

86%

Does it follow best practices?

Impact

89%

1.45x

Average score across 13 eval scenarios

SecuritybySnyk

Advisory

Suggest reviewing before use

Overview
Quality
Evals
Security
Files

task.mdevals/scenario-12/

Target Selection for OrbitLabs

Problem/Feature Description

The discovery pipeline has been run multiple times for OrbitLabs, an observability infrastructure company. Historical run outputs are stored under inputs/runs/ in timestamp-named directories. A product team wants to move through the human-gated selection step using the most up-to-date discovery data available.

Run the target selection workflow for the company slug orbitlabs. Use the inputs directory as the working root when searching for discovery runs. Produce a ranked candidates file showing what options were presented to the decision-maker.

After reviewing the candidates, the product team has decided that none of the available targets are the right fit for the next quarter — they are passing on all of them due to competing roadmap priorities. Record a skip decision with the reason: "Roadmap conflict — all targets deprioritized for Q1 planning cycle."

Complete the full target selection workflow through to persisting and validating the skip decision.

Output Specification

  • candidates-report.md — a markdown file showing the ranked candidates that were presented (the table from Step 3 of the workflow)
  • The selection artifact should be written to the same directory as the chosen discovery file
  • The validator script output should be visible in your working notes or a log file

evals

discovery-output-contract.md

README.md

tile.json