Skip to content

AGENTS.md

Latest commit

 

History

History
201 lines (142 loc) · 6.43 KB

AGENTS.md

File metadata and controls

201 lines (142 loc) · 6.43 KB

AGENTS.md - Depression-Sensitive Web Content Support (DS-WCS)

Overview

This document provides context for OpenCode agents using the Depression-Sensitive Web Content Support (DS-WCS) skill. It defines when agents should apply this skill, how to integrate it with other skills, and example invocation patterns.

What This Skill Enables

The DS-WCS skill enables OpenCode agents to:

  1. Audit web content for depression-sensitive UX patterns
  2. Rewrite problematic content with standards-based rationale
  3. Generate traceability reports mapping recommendations to WCAG 2.2, W3C COGA, ISO 9241-110, and ISO/IEC 30071-1
  4. Provide severity-based prioritization for findings (HIGH/MEDIUM/LOW)

Target Audience

This skill serves:

  • Content designers creating or reviewing interface text
  • UX writers developing error messages, CTAs, and help content
  • Product teams building user-facing interfaces
  • Accessibility specialists expanding beyond technical compliance
  • Inclusive design practitioners implementing cognitive accessibility

When Agents Should Auto-Invoke This Skill

Agents should automatically load and apply this skill when users:

Explicit Triggers

  • Ask to "audit content for depression-sensitive issues"
  • Request "cognitive accessibility review" of content
  • Ask to "review for emotional safety" or "supportive tone"
  • Request "shame-free language" or "reduce cognitive load" improvements
  • Ask about "mental health sensitivity" in interfaces
  • Request review of "error messages," "empty states," "CTAs," or "form validation"

Contextual Triggers

  • Content involves error messages, form validation, or account management
  • Content is for sensitive services (healthcare, mental health, financial, government)
  • User requests "inclusive design" or "accessible content" review
  • Reviewing onboarding flows, progressive disclosure, or help text

Combination Triggers

  • "Audit and rewrite error messages for accessibility"
  • "Review this checkout flow for cognitive accessibility issues"
  • "Make this form content more supportive"

When NOT to Auto-Invoke

Do not invoke this skill when:

  • User requests medical or clinical assessment
  • Request is about therapy or counseling content
  • Simple grammar/typo fixes without accessibility scope
  • Non-user-facing content (code comments, internal documentation)
  • User explicitly requests only technical WCAG compliance

Integration with Other Skills

DS-WCS is complementary to other OpenCode skills:

Accessibility Compliance

  • Skill: accessibility-compliance-accessibility-audit
  • Focus: Technical WCAG compliance (perceivable, operable, understandable, robust)
  • DS-WCS Integration: Adds emotional and cognitive dimensions to accessibility reviews

Web Design Guidelines

  • Skill: web-design-guidelines
  • Focus: Structural and interaction design best practices
  • DS-WCS Integration: Extends guidelines to include content tone and emotional safety

UI/UX Pro Max

  • Skill: ui-ux-pro-max
  • Focus: Comprehensive UI/UX design patterns, color, typography
  • DS-WCS Integration: Adds content-focused UX patterns for emotional safety

Recommended Combinations

For comprehensive UX reviews, agents should consider combining DS-WCS with:

  1. Technical + Emotional Accessibility:

    accessibility-compliance + depression-sensitive-web-content

  2. Design System + Content Review:

    ui-ux-pro-max + depression-sensitive-web-content

  3. Full Accessibility Coverage:

    accessibility-compliance + web-design-guidelines + depression-sensitive-web-content

Example Invocation Patterns

Example 1: Audit Request

User: "Audit src/components/ErrorBoundary.tsx for depression-sensitive content issues"

Agent Response:
[Loads depression-sensitive-web-content skill]
[Reads src/components/ErrorBoundary.tsx]
[Applies audit checklist from resources/implementation-guide.md]
[Generates findings report with severity + standards]

## Findings

### HIGH Severity
- src/components/ErrorBoundary.tsx:15 - Blame language in error message
  - Original: "You broke something. This page doesn't exist."
  - Standards: WCAG 3.3.1, COGA Objective 2, ISO 9241-110 5.7

### MEDIUM Severity
- src/components/ErrorBoundary.tsx:18 - Missing recovery path
  - Original: "Error occurred"
  - Standards: WCAG 3.3.3, COGA Objective 2

## Recommended Rewrites
[See full output format in SKILL.md]

Example 2: Rewrite Request

User: "Rewrite this error message to be more supportive: 'Invalid email! Try again.'"

Agent Response:
[Loads depression-sensitive-web-content skill]
[Consults rewrite library]
[Applies Principle 1: Remove Shame/Blame Language]
[Applies Principle 4: Error Recovery Templates]

## Recommended Rewrite

**Before:**
"Invalid email! Try again."

**After:**
"Email format not recognized.

Please enter an email address in this format: name@domain.com"

**Rationale:**
- Removed blame attribution ("You did wrong")
- Removed exclamation point (calm tone)
- Added concrete example (reduces memory load)
- Provided recovery path

**Standards Mapping:**
- WCAG 3.3.1 (Error Identification)
- WCAG 3.3.3 (Error Suggestion)
- W3C COGA Objective 2: Help users avoid mistakes
- ISO 9241-110: Error tolerance

Example 3: Comprehensive Review

User: "Review the onboarding flow for cognitive accessibility issues"

Agent Response:
[Loads depression-sensitive-web-content skill]
[Reads onboarding flow files]
[Applies audit checklist across all categories:
  - CTAs (urgency, generic labels)
  - Empty states (obligation language)
  - Forms (memory reliance)
  - Progress indicators
  - Help text]
[Generates prioritized findings with standards]
[Provides rewrite recommendations]

Output Quality Standards

When applying this skill, agents should:

  1. Always map to standards - Every finding should reference relevant WCAG, COGA, ISO criteria
  2. Use consistent formatting - Follow the output template in SKILL.md
  3. Prioritize by severity - HIGH findings first, with clear rationale
  4. Provide actionable rewrites - Include specific before/after examples
  5. Maintain scope - Don't provide medical advice; stay in UX/content domain

Resource References

  • SKILL.md: Primary skill definition with workflow
  • resources/implementation-guide.md: Detailed rewrite library, standards matrix, and audit checklist

This document provides agent context. For human-facing documentation, see README.md.