Execute a Bounded Pipeline

Failure pattern

The coding agent decides its own workflow: inspect random files, edit code, run a broad test command, summarize confidently, and stop. If the baseline was already broken or the target failure was never reproduced, the final answer still sounds complete.

Reproduce the failure

await agent.prompt("Fix invite acceptance and tell me when it is done.").send();

There is no forced order: work surface, preflight, reproduce failure, patch, verify, review packet.

Successful Anvia pattern

Use PipelineBuilder to make the coding workflow explicit.

import { PipelineBuilder } from "@anvia/core";

const codingPipeline = new PipelineBuilder<CodingTask>({
  id: "invite-acceptance-fix",
  name: "Invite acceptance fix pipeline",
})
  .step(validateWorkSurface, { name: "Validate work surface" })
  .step(runRepoPreflight, { name: "Run repo preflight" })
  .step(reproduceTargetFailure, { name: "Reproduce target failure" })
  .prompt(codingAgent, { name: "Prepare bounded patch" })
  .step(runRequiredEvidence, { name: "Run required evidence" })
  .step(buildReviewPacket, { name: "Build review packet" })
  .build();

const result = await codingPipeline.run(task);

Why it succeeds

The pipeline prevents the common bad sequence: edit first, understand later. The agent cannot skip the target failure. It cannot treat a broad summary as evidence. The review packet is downstream of verification.

Successful pattern

task
-> validate work surface
-> preflight repo
-> reproduce target failure
-> prepare bounded patch
-> run required checks
-> build review packet

Success check

Each stage should answer one question:

Stage	Question
work surface	what behavior is active?
preflight	is the repo in a known state?
reproduce	can we see the failure before patch?
patch	did the agent stay in scope?
evidence	what proves the behavior changed?
review packet	what should a human inspect?

Next move

Once execution is explicit, record what happened.