Repository navigation
Measure which core-tier procedures most projects need #2129
Description
Part of #2124.
The five core templates load into every chain, at about 1,150 lines measured on 2026-10-22. Several sections are procedures for rare events or one ecosystem. Output: a decision per candidate.
| Candidate | Where |
|---|---|
| Container e2e: disable the resource reaper on rootless runtimes | core/testing.md:272 |
| In-process server on port 0 | core/testing.md:260 |
| Test tiers by directory, one collection hook | core/testing.md:282 |
| Refactor fingerprint, AST equivalence, contract tests | core/testing.md:208-247 |
| Hierarchies: variadic overrides, strict override checker | core/quality.md:90 |
| Module split procedure | core/quality.md:136-148 |
| Citing the software | core/docs.md:169 |
Done when
- Each candidate is kept, moved to an opt-in layer, or moved to the playbook
- Each move measures chain reach first, as ADR-028 requires
Activity
braboj commentedon Oct 22, 2026
A second agent review (2026-10-22) names the same candidates: Python override rules, numerical-output fingerprints and the container-reaper setting. Its framing for the decision: keep the underlying safeguard in core, and load the implementation procedure only when the task applies.