remove-mlinter-rule

name	remove-mlinter-rule
description	Retire a TRF rule from the mlinter. Marks the rule deprecated in rules.toml, deletes its module and tests, and records the removal in the CHANGELOG. Use when asked to remove, retire, drop, or delete a rule.

Remove Mlinter Rule

Input

• <rule id>: the rule to retire, e.g. TRF054. Accept a bare number (54) and normalise it to TRFXXX, zero-padded to three digits.
• Optional: the reason for the removal. Ask for one if none is given — it becomes the CHANGELOG entry and the tombstone description, and it is the only record contributors will have of why the rule went away.

Why a tombstone and not a deletion

Rule ids are referenced from outside this repo: # trf-ignore: TRFXXX comments in the transformers tree, CI configs, and copies of rules.toml passed via --rules-toml. So a removal keeps a tombstone entry in rules.toml instead of deleting the table:

• The engine drops a deprecated id from TRF_RULES, TRF_RULE_SPECS, TRF_RULE_CHECKS, DEFAULT_ENABLED_TRF_RULES, the mlinter.TRFXXX public constants, and --list-rules. Projects that still suppress or configure the id are silently unaffected.
• The docs site is the exception: it keeps publishing a page for the id, built from the tombstone and listed under Removed rules on the rule index. A number that used to fire has to stay findable by whoever meets it in an old CI log.
• Asking for it explicitly (--enable-rules TRFXXX, --rule TRFXXX) fails with exit code 2.
• A rules TOML that still lists the id as an active rule fails the whole run with exit code 2. The bundled rules.toml is the authority here, so this holds for custom files too.
• Leaving trfXXX.py on disk while the id is deprecated is also an error, which is what makes a half-finished removal impossible to ship.

Never reuse a retired number for a new rule: the tombstone stays forever, and add-mlinter-rule picks the next number after the highest trf*.py.

Workflow

1. Confirm the rule exists and read it.

   • grep -n "\[rules.TRFXXX\]" mlinter/rules.toml and read the whole table plus mlinter/trfXXX.py.
   • If the id is already marked deprecated = true, stop and report that it is already retired.
   • Summarise for the user what the rule checks and any allowlist_models / cutoff_date it carries, so they can confirm this is the rule they mean before anything is deleted.

2. Replace the TOML table with a tombstone in mlinter/rules.toml.

   • Keep the table in place at its original position — ordering in this file is by rule id.
   • Delete default_enabled, allowlist_models, cutoff_date, and the whole [rules.TRFXXX.explanation] table.
   • Leave exactly:

     [rules.TRFXXX]
     deprecated = true
     description = "Removed in <version>: <one-line reason>."
     

   • <version> is the version under development in pyproject.toml. The description is published: it is the whole of the retired rule's docs page and its row in the Removed rules table, so write it for someone who just hit the id in a CI log, not as an internal note.

3. Delete the rule module.

   git rm mlinter/trfXXX.py
   

4. Delete the tests.

   • Delete the rule's own test file, tests/test_trfXXX.py.
   • Remove the id from the public-API tests: the assertEqual(public_api.TRFXXX, "TRFXXX") line and the assertIn("TRFXXX", public_api.__all__) line.
   • grep -rn "TRFXXX" tests/ mlinter/ must come back empty except for the tombstone in rules.toml. Watch for helper fixtures or shared source strings that only that rule used, and for any other rule's test that happened to reference it.
   • No new test is needed for the removal itself: test_bundled_deprecated_rules_are_fully_retired loops over every tombstone in the bundled rules.toml and asserts the module and the public constant are gone, so it starts covering the new id automatically.

Reference

• Rule metadata and tombstones: mlinter/rules.toml
• Deprecation handling: _load_rule_specs, _build_rule_checks, _validate_rule_ids, and BUNDLED_DEPRECATED_TRF_RULES in mlinter/mlinter.py
• Public API surface: mlinter/__init__.py
• Rule tests: tests/test_trfXXX.py
• General linter tests: tests/test_mlinter.py (engine behaviour under "Deprecated rules")
• The inverse skill: .ai/skills/add-mlinter-rule/SKILL.md