abilityai/trinity-skills/self-diagnostic

Claude Code and agent skill by abilityai/trinity-skills

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Run self-diagnostics on the agent to verify skills, commands, agents, and dependencies are working. Use when troubleshooting, after configuration changes, or for regular health checks.

Install

npx vskill@latest install abilityai/trinity-skills/self-diagnostic

SKILL.md

Agent Self-Diagnostic

Run comprehensive diagnostics to verify all agent components are functioning correctly.

Quick Start

Run full diagnostics:

/self-diagnostic

Run specific test category:

/self-diagnostic skills-only
/self-diagnostic commands-only
/self-diagnostic agents-only
/self-diagnostic dependencies-only

Arguments

$ARGUMENTS

Safety Principles

CRITICAL: These diagnostics are READ-ONLY and NEVER modify data.

  • All tests use read-only operations (list, read, check)
  • NO file creation during tests
  • NO API calls to paid services
  • Test outputs go to /tmp/agent-diagnostic/ only

Diagnostic Categories

1. Infrastructure Tests

Verify core directories exist:

bash
# Test 1.1: Check .claude directory exists
AGENT_ROOT="${AGENT_ROOT:-.}"
test -d "$AGENT_ROOT/.claude" && echo "PASS: .claude directory exists" || echo "FAIL: .claude directory missing"
# Test 1.2: Check skills directory
test -d "$AGENT_ROOT/.claude/skills" && echo "PASS: skills directory exists" || echo "FAIL: skills directory missing"
# Test 1.3: Check commands directory
test -d "$AGENT_ROOT/.claude/commands" && echo "PASS: commands directory exists" || echo "INFO: commands directory not present"
# Test 1.4: Check agents directory
test -d "$AGENT_ROOT/.claude/agents" && echo "PASS: agents directory exists" || echo "INFO: agents directory not present"
2. Dependency Tests

Verify external dependencies are installed:

bash
# Test 2.1: Python available
python3 --version 2>/dev/null && echo "PASS: Python3 available" || echo "FAIL: Python3 not found"
# Test 2.2: Node.js available (optional)
node --version 2>/dev/null && echo "PASS: Node.js available" || echo "INFO: Node.js not found"
# Test 2.3: uv package manager (optional)
uv --version 2>/dev/null && echo "PASS: uv available" || echo "INFO: uv not found (install: curl -LsSf https://astral.sh/uv/install.sh | sh)"
# Test 2.4: Git available
git --version 2>/dev/null && echo "PASS: Git available" || echo "FAIL: Git not found"
3. Skills Validation Tests

Verify skills are properly configured:

bash
AGENT_ROOT="${AGENT_ROOT:-.}"
# Test 3.1: List all skills and validate structure
for skill_dir in "$AGENT_ROOT"/.claude/skills/*/; do
if [ -d "$skill_dir" ]; then
skill_name=$(basename "$skill_dir")
if [ -f "${skill_dir}SKILL.md" ]; then
# Check for required frontmatter
if head -1 "${skill_dir}SKILL.md" | grep -q "^---"; then
if grep -q "^name:" "${skill_dir}SKILL.md" && grep -q "^description:" "${skill_dir}SKILL.md"; then
echo "PASS: Skill '$skill_name' - valid metadata"
else
echo "FAIL: Skill '$skill_name' - missing name or description"
fi
else
echo "FAIL: Skill '$skill_name' - missing frontmatter"
fi
else
echo "FAIL: Skill '$skill_name' - missing SKILL.md"
fi
fi
done
4. Commands Validation Tests

Verify commands are properly configured:

bash
AGENT_ROOT="${AGENT_ROOT:-.}"
# Test 4.1: List all commands and validate structure
if [ -d "$AGENT_ROOT/.claude/commands" ]; then
for cmd_file in "$AGENT_ROOT"/.claude/commands/*.md; do
if [ -f "$cmd_file" ]; then
cmd_name=$(basename "$cmd_file" .md)
if head -1 "$cmd_file" | grep -q "^---"; then
if grep -q "^description:" "$cmd_file"; then
echo "PASS: Command '/$cmd_name' - valid metadata"
else
echo "FAIL: Command '/$cmd_name' - missing description"
fi
else
echo "FAIL: Command '/$cmd_name' - missing frontmatter"
fi
fi
done
else
echo "INFO: No commands directory found"
fi
5. Agents Validation Tests

Verify sub-agents are properly configured:

bash
AGENT_ROOT="${AGENT_ROOT:-.}"
# Test 5.1: List all agents and validate structure
if [ -d "$AGENT_ROOT/.claude/agents" ]; then
for agent_file in "$AGENT_ROOT"/.claude/agents/*.md; do
if [ -f "$agent_file" ]; then
agent_name=$(basename "$agent_file" .md)
if head -1 "$agent_file" | grep -q "^---"; then
if grep -q "^name:" "$agent_file" || grep -q "^description:" "$agent_file"; then
echo "PASS: Agent '$agent_name' - valid metadata"
else
echo "FAIL: Agent '$agent_name' - missing name or description"
fi
else
echo "FAIL: Agent '$agent_name' - missing frontmatter"
fi
fi
done
else
echo "INFO: No agents directory found"
fi

Output Format

Present results as:

markdown
# Agent Self-Diagnostic Report
**Run Date:** [current date/time]
**Agent Root:** [path]
## Summary
| Category | Passed | Failed | Info |
|----------|--------|--------|------|
| Infrastructure | X | X | X |
| Dependencies | X | X | X |
| Skills | X | X | X |
| Commands | X | X | X |
| Agents | X | X | X |
## Detailed Results
### Infrastructure Tests
[Results...]
### Dependency Tests
[Results...]
### Skills Validation
[Results...]
### Commands Validation
[Results...]
### Agents Validation
[Results...]
## Recommendations
[Any issues found and suggested fixes]

Argument Handling

Based on $ARGUMENTS:

  • No arguments or "full": Run all tests
  • "skills-only": Only test skills validation
  • "commands-only": Only test commands validation
  • "agents-only": Only test agents validation
  • "dependencies-only": Only test infrastructure + dependencies
  • "quick": Only run validation tests (skip functional)

Extending This Diagnostic

To add custom tests for your agent:

  1. Create a custom-tests.sh in this skill directory
  2. The diagnostic will source and run it if present
  3. Use the same PASS/FAIL/INFO output format

Example custom test:

bash
# custom-tests.sh
# Test custom integration
if curl -s --max-time 5 "https://your-api.com/health" | grep -q "ok"; then
echo "PASS: Custom API reachable"
else
echo "FAIL: Custom API unreachable"
fi

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository forks
1
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/abilityai/trinity-skills/self-diagnostic/badge)](https://verified-skill.com/skills/abilityai/trinity-skills/self-diagnostic)

Something wrong with this skill or its page? Tell us.