feat(agents): add skill authoring standard

[bfullam]

Description

Adds a canonical repo-local skill authoring standard directly in .agents/skills/repo-skill-authoring/SKILL.md so agents can create and update MetaMask Mobile repo-local skills with a minimal file set.

This PR introduces the repo-skill-authoring skill as the canonical artifact in .agents/skills/, keeps a thin Claude skill shim that points directly to it, and removes the redundant docs SSOT, Cursor command shim, Claude command shim, and per-skill AGENTS.md entry.

Changelog

CHANGELOG entry: null

Related issues

Fixes:

Manual testing steps

N/A

Screenshots/Recordings

Before

N/A

After

N/A

Pre-merge author checklist

• I've followed MetaMask Contributor Docs and MetaMask Mobile Coding Standards.
• I've completed the PR template to the best of my ability
• I've included tests if applicable
• I've documented my code using JSDoc format if applicable
• I've applied the right labels on the PR (see labeling guidelines). Not required for external contributors.

Pre-merge reviewer checklist

• I've manually tested the PR (e.g. pull and build branch, run the app, test code being changed).
• I confirm that this PR addresses all acceptance criteria described in the ticket it closes and includes the necessary testing evidence such as recordings and or screenshots.

---

Note

Low Risk
Low risk documentation/config-only change that adds new agent skill guidance without impacting runtime app code.

Overview
Adds a new canonical repo-local agent skill, repo-skill-authoring, defining the standard structure, naming rules, and authoring workflow for skills under .agents/skills/.

Registers the skill for the OpenAI/Codex harness via agents/openai.yaml and adds a minimal Claude shim in .claude/skills/ that points back to the canonical .agents SKILL.md.

^{Written by Cursor Bugbot for commit 4cc1eb1. This will update automatically on new commits. Configure here.}

[github-actions]

CLA Signature Action: All authors have signed the CLA. You may need to manually re-run the blocking PR check if it doesn't pass in a few minutes.

[NicolasMassart]

After reviewing the full file set, my main concern is the signal-to-noise ratio: 6 files added, 5 of which are one-liner pointers to the single docs/repo-skill-authoring.md that holds the actual content.

A leaner approach:

• Put content directly in .agents/skills/repo-skill-authoring/SKILL.md — Cursor and Codex already discover skills from there. No need for a separate docs/ file.
• Auto-generate .claude/skills/ and .claude/commands/ shims via a yarn claude:generate:skills script that scans .agents/skills/*/SKILL.md and emits pointers. .gitignore the generated files so they don't pollute the repo.
• Drop .cursor/commands/create-repo-skill.md — Cursor discovers skills from .agents/skills/, not from .cursor/commands/.
• Drop the per-skill AGENTS.md section — skill discovery is already automated by each harness. AGENTS.md is useful for things that aren't auto-discovered (setup, architecture, conventions).

This brings the per-skill cost down from 6 hand-maintained files to 2 (SKILL.md + openai.yaml), with Claude shims derived automatically.

The authoring standard itself is valuable — I'd just inline it into the skill's SKILL.md rather than giving it its own SSOT + pointer chain.

[NicolasMassart]

.cursor/commands/create-repo-skill.md is a double indirection

The Cursor command shim references the Claude command, which itself references the docs SSOT. That's two hops to reach the actual content: Cursor command → Claude command → docs/repo-skill-authoring.md. Every other shim goes directly to the docs. This one should too for consistency.

[NicolasMassart]

Skill discovery is already automated by each harness — Cursor reads available_skills, Claude scans .claude/skills/, Codex scans .agents/skills/. No agent reads AGENTS.md to find skills.

AGENTS.md is useful for things that aren't auto-discovered: project setup commands, architecture overview, testing conventions, directory structure. Adding per-skill entries to it is redundant with the discovery mechanism and creates one more place to keep in sync.

So in my proposed structure, AGENTS.md stays as-is — no per-skill section added.

[github-actions]

🔍 Smart E2E Test Selection

• Selected E2E tags: None (no tests recommended)
• Selected Performance tags: None (no tests recommended)
• Risk Level: low
• AI Confidence: 99%

click to see 🤖 AI reasoning details

E2E Test Selection:
All 3 changed files are purely AI agent skill documentation and configuration files located in .agents/skills/repo-skill-authoring/ and .claude/skills/repo-skill-authoring/. These files define how AI agents should author repo-local skills and contain no application code, test infrastructure, UI components, controllers, navigation, or any other code that could affect the MetaMask Mobile app behavior. There is zero impact on any E2E test flows, and no performance-sensitive code was modified.

Performance Test Selection:
No application code, UI components, controllers, or performance-sensitive code was modified. All changes are purely AI agent skill documentation files with no impact on app performance.

View GitHub Actions results

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud