refactor: reduce low-value Hermes review findings #8

Merged
speedi57 merged 5 commits from ruleset-quality-calibration into main 2026-09-04 09:49:05 +00:00
Owner

Goal

Continue the Hermes ruleset calibration by removing low-value/linter-level findings and making architectural conventions explicitly project-context dependent.

Changes

  • remove the Playwright-specific hardcoded-credential rule because the generic secret rule already covers committed credentials and the Hermes deduplication contract should avoid overlapping findings;
  • remove the unused Python imports/variables rule and delegate this deterministic concern to Ruff/Flake8/other linters;
  • remove the generic print()-to-logger rule because it is primarily a project/style convention rather than a semantic defect;
  • rewrite the React direct-HTTP rule so it only applies when the repository visibly establishes a shared API/data-fetching layer;
  • make that React rule diff-scoped and low severity;
  • explicitly suppress it when no shared API-layer convention exists;
  • bump ruleset version to 1.3.0 and update rule count from 50 to 47.

Why

Hermes should spend review budget on semantic security, correctness, reliability and meaningful performance defects. Deterministic lint/style issues belong to dedicated tooling, while architectural preferences should only be raised when repository/project context proves the convention exists.

Follow-up

After this calibration PR, add semantic fixture tests to exercise representative positive, negative and ambiguous cases and protect the ruleset against false-positive regressions.

## Goal Continue the Hermes ruleset calibration by removing low-value/linter-level findings and making architectural conventions explicitly project-context dependent. ## Changes - remove the Playwright-specific hardcoded-credential rule because the generic secret rule already covers committed credentials and the Hermes deduplication contract should avoid overlapping findings; - remove the unused Python imports/variables rule and delegate this deterministic concern to Ruff/Flake8/other linters; - remove the generic `print()`-to-logger rule because it is primarily a project/style convention rather than a semantic defect; - rewrite the React direct-HTTP rule so it only applies when the repository visibly establishes a shared API/data-fetching layer; - make that React rule diff-scoped and `low` severity; - explicitly suppress it when no shared API-layer convention exists; - bump ruleset version to 1.3.0 and update rule count from 50 to 47. ## Why Hermes should spend review budget on semantic security, correctness, reliability and meaningful performance defects. Deterministic lint/style issues belong to dedicated tooling, while architectural preferences should only be raised when repository/project context proves the convention exists. ## Follow-up After this calibration PR, add semantic fixture tests to exercise representative positive, negative and ambiguous cases and protect the ruleset against false-positive regressions.
speedi57 deleted branch ruleset-quality-calibration 2026-09-04 09:49:09 +00:00
Sign in to join this conversation.
No description provided.