Ten distinct tasks · Agentic coding
Work with a coding agent
Task briefs, repository context, review, refactoring and reversible edits. Choose a specific task for its worked example, adaptable agent brief and acceptance checks.
10 guides found
Write an agentic coding task brief that produces a reviewable change
Describe the observed problem, the desired behavior and the evidence that would establish completion. A worked example, agent brief, failure checks and acceptance evidence.
Map a codebase with AI before editing: entry points, ownership and dependencies
Ask the agent to trace one real user journey from entry point to storage or external service. A worked example, agent brief, failure checks and acceptance evidence.
Plan before agentic code changes: when a short design step saves rework
Use a planning step when a request changes ownership, data shape or more than one subsystem. A worked example, agent brief, failure checks and acceptance evidence.
Keep AI code changes small: reduce review cost without hiding necessary work
Anchor the request to one behavior and keep unrelated cleanup out of the patch. A worked example, agent brief, failure checks and acceptance evidence.
Refactor legacy code with AI: preserve behavior while changing structure
Capture the behavior that must remain stable, then refactor one boundary at a time. A worked example, agent brief, failure checks and acceptance evidence.
Write a bug report for an AI coding agent: reproduce, isolate and verify
Provide the exact action sequence, expected result, observed result and relevant environment. A worked example, agent brief, failure checks and acceptance evidence.
Review an AI-generated pull request: intent, side effects and release evidence
Review the change against its original intent, inspect changed contracts and side effects, and evaluate tests for meaningful coverage. A worked example, agent brief, failure checks and acceptance evidence.
Write repository instructions for AI agents: commands, boundaries and local conventions
Document the project’s actual run and verification commands, source ownership and editing boundaries. A worked example, agent brief, failure checks and acceptance evidence.
Undo AI code changes safely: snapshots, later edits and working copies
Restore from a known snapshot or version-control revision and check whether newer work touches the same files. A worked example, agent brief, failure checks and acceptance evidence.
Use AI across monorepo packages: contracts before cross-package edits
Identify the package that owns the feature and the shared contracts its callers use. A worked example, agent brief, failure checks and acceptance evidence.