pre-commit yamllint hook passes no --config, so local and CI disagree by construction

[twistedmelonman]

Parallel to #531 (zizmor), same defect, same fix shape. Filed after surveying every hook in ~/.config/pre-commit/config.yaml.

Problem

~/.config/pre-commit/config.yaml:91-97:

- id: yamllint
  name: "YAML Lint (yamllint)"
  entry: yamllint
  language: system
  types: [yaml]
  pass_filenames: true

Bare entry: yamllint, no --config. yamllint falls back to its own discovery — a repo-root .yamllint*, else $XDG_CONFIG_HOME/yamllint/config, else its built-in default preset.

CI does not do this. standards/run-standards.sh:146-150:

cfg=""
for c in .yamllint .yamllint.yml .yamllint.yaml; do
  if [[ -f "${repo}/${c}" ]]; then cfg="${repo}/${c}"; break; fi
done
[[ -n "${cfg}" ]] || cfg="${config_dir}/yamllint.yml"

Repo config wins, else the canonical github-workflows/standards/yamllint.yml. So for any repo without its own .yamllint*, pre-commit and CI lint the same file against different rules.

Why it matters

Same class as #531: a local gate that disagrees with CI is not a gate. It fails in both directions — a local pass that CI rejects (wasted push, which Protocol 4 exists to prevent), or a local block on a rule CI does not enforce (which trains SKIP=yamllint as a habit).

Lower severity than #531 in one respect: yamllint's built-in default is stricter than the canonical config on line length, so the common failure is a false block rather than a missed finding. Still wrong, and the fix is identical.

Fix

Per the option-2 decision on #531: github-workflows stays canonical, dotfiles vendors standards/yamllint.yml to a stable local path at install time, and this hook passes it as the fallback when the repo has no config of its own.

Do not copy markdownlint's either/or branch — see #532 for why that shape is itself buggy.

Related: #531 (zizmor, same fix), #532 (markdownlint, the bug not to reproduce), smartwatermelon/dev-env#77.

https://claude.ai/code/session_014Xei4nFGe48rE4vryi64Yx

[twistedmelonman]

Correction: the two references above to "#532" for the markdownlint either/or bug should read #533. #532 is this issue. The markdownlint defect is filed at #533 (and against the dotfiles copy at twistedmelonman/dotfiles#308).