[BRE-1830] feat(bitwarden-devops-engineer): Add naming conventions - #192
Conversation
Add a `workflow-naming-conventions` reference skill covering the GitHub Actions naming standards that `bwwl` does not enforce, so audits stop relying on undocumented convention. The skill defers to `bitwarden-workflow-linter-rules` for every linted rule to avoid double-reporting, and treats job ID and file renames as report-only because both are addressable identifiers. * Document job IDs (kebab-case), step names (Sentence case with a leading imperative verb), and workflow file names (kebab-case.yml, `_` prefix when exclusively reusable) * Add reference-sweep procedures for job ID and filename renames, plus an advisory canonical step-name glossary * Bump plugin to 0.2.0 with changelog entry and README listing * Add `greppable` to the cspell dictionary
Claude Configuration ValidationPR: #192 · Plugin: Result: PASS — no errors. 2 warnings worth fixing before merge, 6 minor/advisory items. Manifest, version consistency across all required files, changelog format, skill rename completeness, and credential scanning all validate clean.
1. Plugin Validation —
|
🤖 Bitwarden Claude Code ReviewOverall Assessment: APPROVE This PR adds a Code Review Details
|
SaintPatrck
left a comment
There was a problem hiding this comment.
Consider having Skill Creator generate evals to provide evidence that all the prose within the new Skill is actually load-bearing, and not overly verbose as Claude likes to do. Without evals, objective review of the Skill content is impossible unless reviewers run and evaluate it manually.
Bring the `workflow-naming-conventions` skill description in line with the skill spec, which forbids XML tags in `description`. Trigger phrasings are folded into prose so the skill still matches the same potential user questions.
Disambiguate this skill's rule IDs from real `bwwl` rule IDs. The three standards are code-fenced headings styled exactly like linter rules, and only kebab-vs-snake casing separates them, so a reader of a generated report could mistake a convention finding for linter output.
Address the ownership line's source-of-truth pointer with a resolvable path instead of a bare skill name, matching the form `action-audit` uses to reach the same skill.
Align the two sentence case headings in `workflow-naming-conventions` with the Title Case used across the plugin's other skill headings. Cosmetic only. * `## Advisory: canonical step names` to `## Advisory: Canonical Step Names` * `## Applying these standards` to `## Applying These Standards`
SaintPatrck
left a comment
There was a problem hiding this comment.
Content looks good. Couple of minor notes about naming and sibling skill references. Nothing worth blocking on though.
…nvocation Point to the linter rules skill by name instead of a file path so the reference works regardless of how the skill is loaded.
Add allowed-tools to workflow-naming-conventions, scoping it to Read, Glob, Grep, Skill, and gh search code. Document in the skill body that it inspects and flags naming deviations but never edits.
Rename `workflow-naming-conventions` to `auditing-workflow-conventions` to better reflect that the skill audits rather than defines conventions. Updates all references in the CHANGELOG, README skill table, and SKILL.md frontmatter.
🎟️ Tracking
BRE-1830
📔 Objective
Add a
workflow-naming-conventionsauditing-workflow-conventionsreference skill covering the GitHub Actions naming standards thatbwwldoes not enforce, so audits stop relying on undocumented convention.The skill defers to
bitwarden-workflow-linter-rulesfor every linted rule to avoid double-reporting, and treats job ID and file renames as report-only because both are addressable identifiers._prefix when exclusively reusable)greppableto the cspell dictionary