I built a workflow that checks whether my AI agent's context was enough, then proves it with my own tests

Description and use

A workflow of Python tools verifies AI agent context sufficiency and validates it with contained tests.

Problem

Uncertainty whether the context provided to an AI coding agent is sufficient for task completion.

Outcome

The workflow identified two verification gaps and a test failure due to a path assumption, and uncovered a bug in the contained runner that hid user site-packages, which was fixed and shipped.

Workflow

• Clone the target repository
• Run ghostworkflow to execute GhostSlice, GhostVerify, GhostBench, and GhostConsole in sequence
• Review the console board for verified status, gaps, and test results
• Address any identified gaps or failures
• Iterate until all checks pass

Evidence

GhostSlice cuts a project down to a task bundle and reports the token cut. On my own bench repo: 74,692 tokens full, 29,926 sliced.

Source

GhostVerify checks the bundle against the task before an agent runs, and says gaps_found when something reachable is missing.

Source

GhostBench runs the same task full vs sliced and, when you give it a test command, executes your tests in a contained runner (scrubbed environment, your secrets never reach the test process) and labels the result honestly: static analysis is not called a live run.

Source

I ran the whole thing on its own repo and published the result warts and all: verdict attention, 2 verify gaps, and a test failure that turned out to be a path assumption in one of my own tests.

Source

The first dogfood run also caught a real bug in the contained runner (it hid the user site-packages, so pytest vanished), which is fixed and shipped.

Source