Add base AGENTS.md/CLAUDE.md agent-instructions system - #378
Conversation
Ports the AGENTS.md + docs/ai/ trigger-based instructions set validated in a client fork back to the base framework, so every future ModernPress project starts with it. Adapted for moose: PHPCS indentation corrected to tabs, and the Tests section rewritten to reference Slic instead of the client fork's lando-composer-test setup. Client-specific hosting config (Dokku hostnames) was dropped rather than ported.
MlKilderkin
left a comment
MlKilderkin
left a comment
There was a problem hiding this comment.
Here are some notes
I also added detailed standards for PHP, comments and PRs in #384
|
|
||
| ## Tests | ||
|
|
||
| PHP tests run via [Slic](https://github.com/stellarwp/slic) (WP-Browser + Codeception in a Docker container). See `docs/php-tests.md` for full setup. |
There was a problem hiding this comment.
In most cases that provides you errors and unexpected behaviour because not everyone has slic installed. AI will think that it needs to install
There was a problem hiding this comment.
@MlKilderkin what would you suggest we do here? Is Slic ever really leveraged in ModernPress projects as suggested in php-tests.md? If not, should we just remove this section and do we need to update the PHP testing for ModernPress altogether?
Add AGENTS.md agent docs with PHP coding standards
|
@LayaTaal child PR is merged and we are ready for review |
dpellenwood
left a comment
dpellenwood
left a comment
There was a problem hiding this comment.
This is awesome. I just left a few Q's around hard-oded php versions which can be messy to remember to update later.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
|
@dpellenwood I updated the docs to check the PHP version instead of hard-coding in the current version. |
jamiepaul
left a comment
jamiepaul
left a comment
There was a problem hiding this comment.
🙌🏻 great work on this @MlKilderkin & @LayaTaal I love how concise the AGENTS.md is after splitting out various additional context files
5 participants
What does this do/fix?
Adds a starting point for layered agent instruction system so AI agents (Claude, Codex, Copilot, Cursor) orient themselves faster and allows designers dabbling in code to get a guided, safer workflow.
CLAUDE.md,AGENTS.md,.github/copilot-instructions.md,.cursor/rules/project.mdc) — ~60 lines each, covering key paths, stack, persona governance, and hard rulesdocs/ai/blocks.md(Gutenberg conventions),docs/ai/php.md(namespaces, DI container, PHPCS),docs/ai/commands.md(lando/npm/wp-cli),docs/ai/designer-mode.md(allowed files, git workflow, regression safety)local-config.json;user_role = designerloads designer-mode restrictions silently