Agents

Agents

An agent is a specialist your AI tool can hand work to — a reviewer, an analyst, a tester. Each has one job and clear limits.

  • AgentUSD 2

    Documentation Engineer

    Writes and repairs the documentation developers actually need — READMEs that get you running, API references that match reality, and architecture notes that explain why — verified against the code, not copied from hope.

    Best for: Projects whose README lies about how to run them

  • AgentUSD 2

    Test Strategist

    Designs the test suite your system actually needs — what to test at which level, where the risk lives, what to do about the flaky tests — and builds the missing pieces, so green means something again.

    Best for: Projects with either no tests or thousands of worthless ones

  • AgentUSD 2

    CI Pipeline Engineer

    Builds and repairs continuous-integration pipelines that are fast, trustworthy and understandable — the right checks in the right order, failures that explain themselves, and no more 'just re-run it'.

    Best for: Teams whose pipeline is slow, flaky or a mystery

  • AgentUSD 2

    Incident Commander

    Runs a production incident with discipline: stabilise first, diagnose with evidence, communicate at a steady cadence, and capture the timeline — so the outage ends sooner and teaches more.

    Best for: Teams with no on-call playbook and occasional chaos

  • AgentUSD 2

    Security Hardener

    Reviews your application the way an attacker reads it — inputs, auth, secrets, dependencies, data exposure — then fixes what it can and hands you a ranked plan for the rest.

    Best for: Products that have never had a security pass

  • AgentUSD 2

    Dependency Steward

    Keeps your dependencies healthy on purpose: audits what you have, upgrades safely in order, removes what you do not need, and tells you clearly when a vulnerability actually matters for you.

    Best for: Projects years behind on upgrades and afraid to start

  • AgentUSD 2

    Data Modeller

    Designs database schemas that stay truthful under load and change: entities that match the business, constraints that make bad data impossible, and the awkward questions asked before the data exists.

    Best for: New features that need tables designed properly the first time

  • AgentUSD 2

    Data Pipeline Architect

    Designs data pipelines that survive reality: idempotent steps, failures that recover without double-counting, late data handled on purpose, and freshness guarantees someone actually wrote down.

    Best for: Replacing the nightly script that silently fails with something trustworthy

  • AgentUSD 2

    Prompt Engineer

    Designs, tests and hardens the prompts your product or workflow runs on — structured instructions, honest evaluation against real cases, and versions you can roll back — instead of vibes in a text box.

    Best for: Products with an LLM feature that works 'usually'

  • AgentUSD 2

    Model Evaluator

    Builds the evaluation harness that tells you whether your AI feature is actually good: real test sets, metrics that match what users care about, and regression gates so quality cannot silently slide.

    Best for: AI features judged today by 'it seems fine'

Also see the Skills.