tools/aissfw.py grew to about 9,500 lines, and the author now questions its boundaries

The article follows an earlier piece that described why ALPACA-template placed framework/catalog.toml as a machine-readable contract separate from human- or AI-facing rules. Here the focus shifts to the tool that reads it: tools/aissfw.py.

The author separates roles deliberately. People, AI, and QA judge meaning and validity, while the CLI handles structure and consistency such as required artifacts, IDs, relationships, table structure, and gate prerequisites. Final decisions stay with QA and the Project Owner. The CLI supports commands like framework check, project check, trace check, trace render, and gate preflight, and it validates the catalog itself before checking any project so a broken rule set cannot reject artifacts. Results are collected into a shared CheckResult and returned as Markdown or JSON with distinct exit codes. Only trace render --write is allowed to write, and the CLI never treats its pass as an approval.

The author is candid that the single-CLI approach has become unwieldy at about 9,500 lines. The stated reason is not the line count itself but that different responsibilities share one module, so tracing the impact of a contract change across the parser, catalog interpretation, gate aggregation, and output format takes mental effort. The next version's rework is framed as drawing boundaries that only became visible after the framework was actually run, which is a reminder that early-stage consolidation can be a reasonable path to clarity rather than purely a liability.