Skills

Skills are reusable procedures that agents load to perform specific tasks. They provide structured approaches and consistent behavior across sessions.

Available Skills

Skill	Used By	Purpose
approach-evaluation	coder, orchestration	Evaluate implementation alternatives before coding
code-review	reviewer	Structured code review with verdicts
code-validation	orchestration	Type check, lint, and test validation
test-generation	test-writer	Framework-aware test writing
systematic-debugging	debugger	Root cause analysis methodology
researching-topics	researcher	Multi-source documentation research
brainstorming	plan	Strategic planning and approach evaluation
docs-generator	docs-generator	Generate /docs for Notion sync
README-generator	readme-generator	Generate project README
directives-advisor	orchestration	Help configure directives
system-discovery	orchestration	Discover project tooling and architecture
project-status	orchestration	Check Era governance and git status
hosted-apps	orchestration	Deploy apps to *.era.computer

Implementation Pipeline Skills

Approach Evaluation

Evaluates 2-3 implementation alternatives before coding non-trivial features.

When used: Before implementing features that touch 3+ files or have multiple valid approaches.

Output:

• 2-3 approaches with descriptions
• Evaluation on: complexity, risk, alignment, testability
• Selected approach with rationale

Skipped when:

• Change is < 20 lines or single file
• User specified the exact approach
• Bug fix with identified root cause

Code Review

Produces structured APPROVE/REJECT verdicts with findings.

Review dimensions:

• Correctness — logic, edge cases, error handling
• Style — naming, formatting, patterns
• Security — secrets, injection, validation
• Tests — coverage, assertions, independence
• Architecture — layers, separation, abstractions

Finding severities:

• BLOCKER — Must fix. Any blocker = REJECT.
• WARNING — Should fix. 3+ warnings = REJECT.
• SUGGESTION — Nice to have. Never causes rejection.

Code Validation

The "Iron Law" — no code returns without validation first.

Phases:

1. Discovery — Detect project tooling (test framework, linter, type checker)
2. Validate — Run type check → lint → targeted tests → full test suite
3. Fix — Address failures (max 3 attempts)
4. Report — Structured pass/fail output

Supported frameworks:

• Jest, Vitest, Playwright, Mocha (JavaScript/TypeScript)
• pytest, unittest (Python)
• cargo test (Rust)
• go test (Go)

Test Generation

Framework-aware test writing that matches existing project patterns.

Process:

1. Detect test framework from existing tests
2. Analyze existing test patterns (describe/it structure, mocking, assertions)
3. Plan coverage (happy path, errors, edge cases)
4. Generate tests matching project style
5. Run tests to verify they pass

Principles:

• Test behavior, not implementation
• One behavior per test
• Minimal mocking (mock at boundaries only)
• Independent tests (no shared state)

Using Skills

Agents load skills automatically based on their instructions. The orchestration agent coordinates which agents (and thus which skills) are invoked based on the task.

Skills are deployed to .era/resources/opencode/skill/ during era-code init and can be customized per project.

Slash Commands

Several slash commands invoke skills directly:

Command	Skill / Agent	Purpose
/era-validate	code-validation	Run validation on project
/era-debug	systematic-debugging	Start debugging session
/era-research	researching-topics	Research a topic
/era-plan	brainstorming	Strategic planning
/era-readme	README-generator	Generate README
/decision-log	orchestration	Log decision + risk assessment to Google Sheets
/linear-create	plan	Create Linear projects or issues from context
/era-deploy	orchestration	Deploy to *.tools.era.computer