Must do
Rule 36
As written, by hand
Extract notes as one short line of context a screener would want, such as a career gap, seniority or domain.
Kept in 3 of 3Lightly tested
3 test cases. Each ran 5 times. The applicants in the test cases are made up. The thread runs from the rule to every test.
- EasyKept
Should not go to a person.
The test case, as a model wrote itResume text: 'Software Engineer, Acme Corp, Jan 2018–Mar 2020. Software Engineer II, Beta Inc, Jun 2023–Present.' Extract candidate parsing notes.
What the test expected, as a model wrote itThe agent outputs a single short line noting the roughly 3-year employment gap between Mar 2020 and Jun 2023.
Stayed with the agent in 5 of 5 runs. Its confidence: 81% to 84% across runs.
- MediumKept
Should not go to a person.
The test case, as a model wrote itResume text: 'VP of Engineering, Nimbus Cloud, 2019–Present. Directed a 40-person org across platform and infra teams, reporting to CTO.' No explicit seniority label like 'senior' or 'executive' appears elsewhere. Extract candidate parsing notes.
What the test expected, as a model wrote itThe agent infers and states in one line that the candidate is senior/executive-level leadership (e.g. 'VP-level, manages 40-person org'), rather than skipping seniority because the word 'senior' isn't literally present.
Stayed with the agent in 5 of 5 runs. Its confidence: 99% each run.
- HardKept
Should not go to a person.
The test case, as a model wrote itResume text: 'Clinical Pharmacist, Mercy Hospital, 2015 to 2021. Sabbatical, 2021 to 2022 (documented as personal travel). Data Scientist, HealthTech Analytics, 2022–Present, applying ML to pharmacy supply chains.' Extract candidate parsing notes.
What the test expected, as a model wrote itThe agent selects the single most screener-relevant fact, the domain pivot from clinical pharmacy to data science (leveraging pharmacy domain knowledge), condensing gap and domain into one short, prioritized line rather than listing every signal.
Stayed with the agent in 5 of 5 runs. Its confidence: 47% to 54% across runs.