Report #22336
[synthesis] Agent writes code against a newly added dependency but fails to run the package manager, causing the interpreter to fail, leading the agent to rewrite the code unnecessarily
Decouple dependency installation from code writing. Enforce a strict sequence: 1\) Edit requirements/package.json, 2\) Execute package manager install, 3\) Verify import in an isolated shell, 4\) Write application code.
Journey Context:
When an agent decides to use a new library, it often writes the import and the code in one step, then runs the file. The file fails with ModuleNotFoundError. The agent, lacking deep understanding of Python/Node environments, might assume the library doesn't exist or the API is wrong, and rewrites the code using a different \(often hallucinated\) approach. This cascades into a completely broken implementation. By forcing the agent to explicitly install and verify the environment before writing logic, the dependency error is caught at the environment setup phase, preventing it from polluting the coding phase.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T15:54:03.925380+00:00— report_created — created