diff --git a/agents/darcyegb__claude-code-agents/README.md b/agents/darcyegb__claude-code-agents/README.md
new file mode 100644
index 0000000..2b075e2
--- /dev/null
+++ b/agents/darcyegb__claude-code-agents/README.md
@@ -0,0 +1,54 @@
+# Claude Code Agents Collection
+
+A curated set of **seven specialized QA and verification sub-agents** for
+[Claude Code](https://claude.ai/code). Each agent owns a distinct quality-assurance
+discipline so your team has the right expert for every review scenario.
+
+## Agents
+
+| Agent | File | What it does |
+|---|---|---|
+| **Jenny** | `Jenny.md` | Compares the live codebase against written specifications, file-by-file, line-by-line. Surfaces gaps with severity levels and file references. |
+| **Karen** | `karen.md` | Goes and *runs* the thing. Validates claimed completions by executing the code path, not by reading it. Blunt about what is and isn't working. |
+| **Claude MD Compliance Checker** | `claude-md-compliance-checker.md` | Ensures every change obeys the project's `CLAUDE.md` conventions. Catches drift before it becomes habit. |
+| **Code Quality Pragmatist** | `code-quality-pragmatist.md` | Hunts over-engineering: enterprise abstractions in MVPs, unnecessary caches, infinite middleware. Pushes toward simple, maintainable code. |
+| **Task Completion Validator** | `task-completion-validator.md` | Verifies "done" means "actually works". Looks for TODOs, empty catch blocks, tests that only test mocks, and fake integrations. |
+| **UI Comprehensive Tester** | `ui-comprehensive-tester.md` | Systematic UI validation across web and mobile. Auto-selects Puppeteer, Playwright, or Mobile MCP. Tests flows and edge cases. |
+| **Ultrathink Debugger** | `ultrathink-debugger.md` | Deep root-cause analysis when other attempts have failed. Traces execution end-to-end, fixes causes not symptoms. |
+
+## How to use
+
+Copy any `.md` file into your project's `.claude/agents/` directory. Claude Code
+will pick it up as a sub-agent. Invoke it by name when you need its specialty.
+
+### Example workflow
+
+```
+# After implementing a feature:
+task-completion-validator → confirm it works
+code-quality-pragmatist   → check for over-engineering
+claude-md-compliance-checker → verify project rules
+```
+
+```
+# When something is "done" but feels off:
+karen → reality-check the completion
+jenny → spec-vs-implementation audit
+```
+
+## Cross-agent collaboration
+
+The agents are designed to delegate to each other:
+- Jenny calls Karen for end-to-end validation
+- Karen calls Jenny when spec complexity is high
+- All agents use `file_path:line_number` format and shared severity levels (`Critical | High | Medium | Low`)
+
+## GAP compatibility
+
+This collection ships with an `agent.yaml` + `SOUL.md` at the repo root, making it
+fully compatible with the [GitAgent Protocol](https://gitagent.sh). Drop it into
+any GAP-compatible runtime and it works out of the box.
+
+## Repository
+
+https://github.com/darcyegb/ClaudeCodeAgents
diff --git a/agents/darcyegb__claude-code-agents/metadata.json b/agents/darcyegb__claude-code-agents/metadata.json
new file mode 100644
index 0000000..891e5fe
--- /dev/null
+++ b/agents/darcyegb__claude-code-agents/metadata.json
@@ -0,0 +1,15 @@
+{
+  "name": "claude-code-agents",
+  "author": "darcyegb",
+  "description": "Seven specialized QA sub-agents for Claude Code: spec verification, reality-check, CLAUDE.md compliance, code-quality pragmatism, task validation, UI testing, and deep debugging.",
+  "repository": "https://github.com/computer-agent/ClaudeCodeAgents",
+  "path": "",
+  "version": "1.0.0",
+  "category": "testing",
+  "tags": ["qa", "code-quality", "verification", "claude-code", "debugging", "compliance", "testing"],
+  "license": "MIT",
+  "model": "claude-sonnet-4-5-20250929",
+  "adapters": ["claude-code", "system-prompt"],
+  "icon": false,
+  "banner": false
+}