Full Output Keeper

When to use

When a build, test, search, or diagnostic command may produce more output than the agent can read at once, use the bundled Python 3 standard-library helper. When earlier output is already lost, rerun only if the command is safe and the user authorized its effects.

Steps

1. Resolve scripts/aunx-keep.py relative to this skill folder. Read the operating reference when setting a cache location or investigating a helper failure.
2. Choose a useful view: --head N for opening context or --tail N for a final summary. Add --max-bytes N when lines may be large. The footer is outside that display budget.
3. Run the authorized command directly after --. Use -c only when shell syntax is needed. Preserve the command's returned status.
4. Read the footer: retrieval id, line range, total size and exit code. If storage failed, report the stated failure and avoid promising full recall.
5. Before making a conclusion about omitted errors or warnings, retrieve the relevant range or search the saved output. Keep the id alongside the result so a later turn can retrieve more.

From the skill folder: