Claude Skill

drupal-ai-contrib

Routes Drupal contribution work to the right contribution-quality stage and supplies the contribution knowledge layer. Use when user says 'contribute to Drupal', 'Drupal contribution', 'submit a Drupal patch', 'Drupal merge request', 'fix a Drupal core issue', 'contribute a modul

LLM Mart · 0 points · 1 views 0 listing impressions 0 install-command copies
Download camoa-claude-skills-drupal-ai-contrib_skills_drupal-ai-contrib-c73ba2a.zip · 7 KB

Install

skills CLI npx skills add https://github.com/camoa/claude-skills/tree/main/drupal-ai-contrib/skills/drupal-ai-contrib
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install camoa-claude-skills@llmmart
Git git clone https://github.com/camoa/claude-skills.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole camoa/claude-skills collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Drupal AI Contribution — Umbrella & Router

Make AI a trusted, good-practice way to contribute to Drupal: AI-assisted code that passes drupalci, survives maintainer review on the first or second round, follows Drupal + PHP best practices, and follows the Code of Conduct and the adopted AI-contribution policy.

This skill routes a contribution request to the right stage and supplies the knowledge layer. It does not run the stages itself — each stage is a worker skill.

Core principle — evidence over assertion

Every gate passes only on a produced artifact: a command's output, a file diff, a live API response, a real pipeline result. Never on the model stating "done", "passes", or "should work". The cross-cutting discipline lives in the contribution-guardrails skill — invoke it during any development between issue and verify.

The contribution arc — detect-driven, not a forced sequence

Six stages. Each checks the state it needs and runs regardless of whether earlier stages ran. A contributor can enter at issue (environment ready) or even at verify (work underway). A stage that finds a real gap points at setup for that gap only; it never refuses to run. setup is the on-ramp, never a prerequisite.

Stage Command Worker skill
1. Onboard the environment /drupal-ai-contrib:setup contribution-setup
2. Work the issue /drupal-ai-contrib:issue contribution-issue
3. Verify locally (the inner loop) /drupal-ai-contrib:verify contribution-verify
4. Honest review /drupal-ai-contrib:review contribution-review
5. Submit the MR /drupal-ai-contrib:submit contribution-submit
6. Confirm the real pipeline /drupal-ai-contrib:pipeline contribution-pipeline

Routing

When a contribution request arrives without a specific command, route it:

The contributor wants to… Route to
Stand up a DDEV environment, scaffold CI config, install Drupal AI skills contribution-setup
Find / create / claim an issue, check out an issue fork + branch contribution-issue
Run the local drupalci-parity / AI-policy / eval gates contribution-verify
Get an honest, fresh-context review of the work contribution-review
Create or update the merge request, write the AI-disclosure comment contribution-submit
Check the authoritative real GitLab MR pipeline contribution-pipeline
Develop code safely between issue and verify contribution-guardrails

Ambiguous request → ask which stage; do not guess.

The knowledge layer — dev-guides

Technical how-to (writing the code correctly, the contribution mechanics, coding standards, the drupalci pipeline) comes from the camoa/dev-guides contribution guides, not from this plugin. Load them via the dev-guides-navigator skill — never fetch llms.txt or dev-guides URLs directly.

references/dev-guides-index.md maps each contribution stage to the dev-guide slugs to load. Worker skills cite those slugs. A guide that is not yet authored simply will not resolve — the navigator degrades gracefully; never block on a missing guide.

The governance layer — ai_best_practices

AI-policy + compliance + evals anchor on ai_best_practices (the canonical Drupal AI source of truth) and the adopted Policy on the use of AI when contributing to Drupal. Both are fetched live per contribution by the ai-policy-checker agent — never hard-coded, never assumed; this is the fastest-moving area.

Relationship to ai-dev-assistant

A contribution is an ai-dev-assistant task. The framework owns the phase lifecycle (research → architecture → implement → review → complete) and its phase gates. This plugin is a separate layer outside the framework — it adds contribution-quality commands and gates on top. It does not fork or modify the framework. Run the contribution as an ai-dev-assistant task and use this plugin's commands alongside it.

Interop — delegate, never reinvent

Need Delegate to
Phase lifecycle + phase gates ai-dev-assistant
Philosophy / standards review (SOLID, DRY) code-quality-tools
Paper-testing before submission code-paper-test
Issue / MR / pipeline CLI mglaman/drupalorg-cli — executable drupalorg, wrapped by issue / submit / pipeline; see references/drupalorg-cli.md
Drupal AI skill install drupal_devkit

Examples

Example 1: natural-language entry, no command

User says: "I want to contribute a fix to the Pathauto module." Actions:

  1. Recognize a Drupal contribution request with no specific stage named.
  2. There is no environment context yet → route to contribution-setup to detect the workflow (someone else's contrib) and environment, then contribution-issue. Result: The contributor is on-ramped at the right stage without guessing.

Example 2: mid-arc entry

User says: "Check my Drupal contribution before I submit it." Actions:

  1. Work is already underway → skip setup/issue.
  2. Route to contribution-verify (gates), then contribution-review (honest review). Result: Entry at verify — earlier stages are not forced.

Troubleshooting

Situation Handling
Request could be two stages Ask which stage; never guess the route.
A dev-guide slug does not resolve The navigator degrades gracefully — proceed without it; never block.
Contributor asks to skip a stage Allowed — the arc is detect-driven. Surface what evidence the skipped stage would have produced.
Request is Drupal dev but not contribution Out of scope — this is contribution quality, not general Drupal dev. Defer to ai-dev-assistant.

References

  • references/dev-guides-index.md — contribution stage → dev-guide slugs to load
  • references/drupalorg-cli.md — the drupalorg CLI: what it is, install, subcommands
Files (claude-skills)
  • references
    • dev-guides-index.md 4.1 KB
      Loading…
    • drupalorg-cli.md 7.5 KB
      Loading…
  • SKILL.md 6.5 KB
    ---
    name: drupal-ai-contrib
    description: "Routes Drupal contribution work to the right contribution-quality stage and supplies the contribution knowledge layer. Use when user says 'contribute to Drupal', 'Drupal contribution', 'submit a Drupal patch', 'Drupal merge request', 'fix a Drupal core issue', 'contribute a module', 'AI-assisted Drupal contribution', or works a drupal.org / GitLab Drupal issue. Use PROACTIVELY whenever AI-assisted code is headed for a Drupal contribution — drupalci and maintainer review are unforgiving."
    version: 0.1.1
    model: inherit
    user-invocable: false
    ---
    
    # Drupal AI Contribution — Umbrella & Router
    
    Make AI a trusted, good-practice way to contribute to Drupal: AI-assisted code that
    passes drupalci, survives maintainer review on the first or second round, follows
    Drupal + PHP best practices, and follows the Code of Conduct and the adopted
    AI-contribution policy.
    
    This skill **routes** a contribution request to the right stage and supplies the
    **knowledge layer**. It does not run the stages itself — each stage is a worker skill.
    
    ## Core principle — evidence over assertion
    
    Every gate passes **only on a produced artifact**: a command's output, a file diff, a
    live API response, a real pipeline result. Never on the model stating "done", "passes",
    or "should work". The cross-cutting discipline lives in the `contribution-guardrails`
    skill — invoke it during any development between `issue` and `verify`.
    
    ## The contribution arc — detect-driven, not a forced sequence
    
    Six stages. Each checks the state it needs and runs regardless of whether earlier
    stages ran. A contributor can enter at `issue` (environment ready) or even at `verify`
    (work underway). A stage that finds a real gap points at `setup` for *that gap only*;
    it never refuses to run. `setup` is the on-ramp, never a prerequisite.
    
    | Stage | Command | Worker skill |
    |-------|---------|--------------|
    | 1. Onboard the environment | `/drupal-ai-contrib:setup` | `contribution-setup` |
    | 2. Work the issue | `/drupal-ai-contrib:issue` | `contribution-issue` |
    | 3. Verify locally (the inner loop) | `/drupal-ai-contrib:verify` | `contribution-verify` |
    | 4. Honest review | `/drupal-ai-contrib:review` | `contribution-review` |
    | 5. Submit the MR | `/drupal-ai-contrib:submit` | `contribution-submit` |
    | 6. Confirm the real pipeline | `/drupal-ai-contrib:pipeline` | `contribution-pipeline` |
    
    ## Routing
    
    When a contribution request arrives without a specific command, route it:
    
    | The contributor wants to… | Route to |
    |---------------------------|----------|
    | Stand up a DDEV environment, scaffold CI config, install Drupal AI skills | `contribution-setup` |
    | Find / create / claim an issue, check out an issue fork + branch | `contribution-issue` |
    | Run the local drupalci-parity / AI-policy / eval gates | `contribution-verify` |
    | Get an honest, fresh-context review of the work | `contribution-review` |
    | Create or update the merge request, write the AI-disclosure comment | `contribution-submit` |
    | Check the authoritative real GitLab MR pipeline | `contribution-pipeline` |
    | Develop code safely between `issue` and `verify` | `contribution-guardrails` |
    
    Ambiguous request → ask which stage; do not guess.
    
    ## The knowledge layer — dev-guides
    
    Technical how-to (writing the code correctly, the contribution mechanics, coding
    standards, the drupalci pipeline) comes from the `camoa/dev-guides` contribution
    guides, **not** from this plugin. Load them via the `dev-guides-navigator` skill —
    never fetch `llms.txt` or dev-guides URLs directly.
    
    `references/dev-guides-index.md` maps each contribution stage to the dev-guide slugs to
    load. Worker skills cite those slugs. A guide that is not yet authored simply will not
    resolve — the navigator degrades gracefully; never block on a missing guide.
    
    ## The governance layer — ai_best_practices
    
    AI-policy + compliance + evals anchor on `ai_best_practices` (the canonical Drupal AI
    source of truth) and the adopted *Policy on the use of AI when contributing to Drupal*.
    Both are fetched **live** per contribution by the `ai-policy-checker` agent — never
    hard-coded, never assumed; this is the fastest-moving area.
    
    ## Relationship to ai-dev-assistant
    
    A contribution **is** an `ai-dev-assistant` task. The framework owns the phase lifecycle
    (research → architecture → implement → review → complete) and its phase gates. This
    plugin is a separate layer **outside** the framework — it adds contribution-quality
    commands and gates on top. It does not fork or modify the framework. Run the contribution
    as an ai-dev-assistant task and use this plugin's commands alongside it.
    
    ## Interop — delegate, never reinvent
    
    | Need | Delegate to |
    |------|-------------|
    | Phase lifecycle + phase gates | `ai-dev-assistant` |
    | Philosophy / standards review (SOLID, DRY) | `code-quality-tools` |
    | Paper-testing before submission | `code-paper-test` |
    | Issue / MR / pipeline CLI | `mglaman/drupalorg-cli` — executable `drupalorg`, wrapped by `issue` / `submit` / `pipeline`; see `references/drupalorg-cli.md` |
    | Drupal AI skill install | `drupal_devkit` |
    
    ## Examples
    
    ### Example 1: natural-language entry, no command
    **User says:** "I want to contribute a fix to the Pathauto module."
    **Actions:**
    1. Recognize a Drupal contribution request with no specific stage named.
    2. There is no environment context yet → route to `contribution-setup` to detect the
       workflow (someone else's contrib) and environment, then `contribution-issue`.
    **Result:** The contributor is on-ramped at the right stage without guessing.
    
    ### Example 2: mid-arc entry
    **User says:** "Check my Drupal contribution before I submit it."
    **Actions:**
    1. Work is already underway → skip `setup`/`issue`.
    2. Route to `contribution-verify` (gates), then `contribution-review` (honest review).
    **Result:** Entry at `verify` — earlier stages are not forced.
    
    ## Troubleshooting
    
    | Situation | Handling |
    |-----------|----------|
    | Request could be two stages | Ask which stage; never guess the route. |
    | A dev-guide slug does not resolve | The navigator degrades gracefully — proceed without it; never block. |
    | Contributor asks to skip a stage | Allowed — the arc is detect-driven. Surface what evidence the skipped stage would have produced. |
    | Request is Drupal dev but not contribution | Out of scope — this is contribution quality, not general Drupal dev. Defer to `ai-dev-assistant`. |
    
    ## References
    
    - `references/dev-guides-index.md` — contribution stage → dev-guide slugs to load
    - `references/drupalorg-cli.md` — the `drupalorg` CLI: what it is, install, subcommands
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.