| name | config-validator |
| description | Self-validation skill for the .agent directory. Checks that all agents, skills, workflows, and scripts referenced across the system actually exist and are consistent. Use after modifying agent configuration files. |
| version | 1.0.0 |
| last-updated | "2026-03-12T00:00:00.000Z" |
| applies-to-model | gemini-2.5-pro, claude-3-7-sonnet |
Config Validator — Agent System Self-Check
This skill validates the internal consistency of the .agent/ directory itself. When the agent system references files that don't exist, behavior becomes unpredictable. This skill catches those gaps.
When to Use
- After adding, renaming, or removing any agent, skill, workflow, or script
- After copying the
.agent/ directory to a new project
- When something "should work" but the agent seems to ignore it
- As part of
/audit to ensure the agent system itself is healthy
What Gets Checked
1. Agent File Existence
Every agent referenced in rules/GEMINI.md routing table must have a corresponding .md file in agents/.
For each agent in the routing table:
→ Does agents/{agent-name}.md exist?
→ If not: report as MISSING AGENT
2. Skill References in Agent Frontmatter
Every skill listed in an agent's skills: frontmatter field must exist as a directory in skills/ with a SKILL.md file.
For each agent file:
→ Read YAML frontmatter
→ For each skill in skills: field
→ Does skills/{skill-name}/SKILL.md exist?
→ If not: report as MISSING SKILL
3. Workflow File Existence
Every slash command listed in GEMINI.md or ARCHITECTURE.md must have a corresponding .md file in workflows/.
For each /command referenced:
→ Does workflows/{command}.md exist?
→ If not: report as MISSING WORKFLOW
4. Script File Existence
Every script referenced in rules/GEMINI.md script table must exist in scripts/.
For each script in the reference table:
→ Does scripts/{script-name} exist?
→ If not: report as MISSING SCRIPT
5. Cross-Reference Consistency
- Agent names in the routing table match filenames in
agents/
- Workflow names in the command table match filenames in
workflows/
- No orphan files (files that exist but are never referenced anywhere)
Validation Process
Run this check manually or mentally when modifying the .agent/ structure:
Step 1: Read rules/GEMINI.md → Extract agent names, script names
Step 2: Read GEMINI.md → Extract slash command names
Step 3: Read ARCHITECTURE.md → Extract all references
Step 4: Read each agent .md → Extract skill references from frontmatter
Step 5: Cross-check every reference against the filesystem
Step 6: Report any mismatches
Report Format
🔧 Config Validation Report
━━━━━━━━━━━━━━━━━━━━━━━━━
Agents: 27 found, 27 referenced ✅
Skills: 37 found, 34 referenced ⚠️ (3 unreferenced)
Workflows: 22 found, 22 referenced ✅
Scripts: 10 found, 10 referenced ✅
Issues:
❌ MISSING: skills/some-removed-skill/SKILL.md (referenced by agents/backend-specialist.md)
⚠️ ORPHAN: agents/old-unused-agent.md (not referenced in routing table)
Fixing Common Issues
| Issue | Fix |
|---|
| Missing agent file | Create the agent .md file or remove from routing table |
| Missing skill directory | Create skills/{name}/SKILL.md or remove from agent skills: field |
| Missing workflow file | Create workflows/{name}.md or remove from slash command table |
| Missing script | Create the script or remove from script reference table |
| Orphan file | Either reference it somewhere or delete it |
Hallucination Guard
- Never report a file as "existing" without actually checking the filesystem
- Never report a reference as "valid" without reading the referencing file
- If a file exists but has different content than expected, flag it rather than assuming it's correct
🤖 LLM-Specific Traps
AI coding assistants often fall into specific bad habits when dealing with this domain. These are strictly forbidden:
- Over-engineering: Proposing complex abstractions or distributed systems when a simpler approach suffices.
- Hallucinated Libraries/Methods: Using non-existent methods or packages. Always
// VERIFY or check package.json / requirements.txt.
- Skipping Edge Cases: Writing the "happy path" and ignoring error handling, timeouts, or data validation.
- Context Amnesia: Forgetting the user's constraints and offering generic advice instead of tailored solutions.
- Silent Degradation: Catching and suppressing errors without logging or re-raising.
🏛️ Tribunal Integration (Anti-Hallucination)
Slash command: /review or /tribunal-full
Active reviewers: logic-reviewer · security-auditor
❌ Forbidden AI Tropes
- Blind Assumptions: Never make an assumption without documenting it clearly with
// VERIFY: [reason].
- Silent Degradation: Catching and suppressing errors without logging or handling.
- Context Amnesia: Forgetting the user's constraints and offering generic advice instead of tailored solutions.
✅ Pre-Flight Self-Audit
Review these questions before confirming output:
✅ Did I rely ONLY on real, verified tools and methods?
✅ Is this solution appropriately scoped to the user's constraints?
✅ Did I handle potential failure modes and edge cases?
✅ Have I avoided generic boilerplate that doesn't add value?
🛑 Verification-Before-Completion (VBC) Protocol
CRITICAL: You must follow a strict "evidence-based closeout" state machine.
- ❌ Forbidden: Declaring a task complete because the output "looks correct."
- ✅ Required: You are explicitly forbidden from finalizing any task without providing concrete evidence (terminal output, passing tests, compile success, or equivalent proof) that your output works as intended.