mikeparcewski/wicked-garden / skills/agentic
wicked-garden-agentic
VerifiedAgentic-systems domain skill: review, design, audit, and framework selection for AI agent systems. Routes to one of four actions (review | design | audit | frameworks) backed by the rubrics in refs/. Use when: "agentic review", "review this agent system", "full agentic codebase review", "trust and safety audit", "audit my agents for GDPR/HIPAA/SOC2/NIST compliance", "design an agentic system", "design a multi-agent architecture", "which agentic framework should I use", "compare LangChain vs CrewAI vs AutoGen", or any former /wicked-garden:agentic:{review|design|audit|frameworks} invocation. Disambiguation: this domain reviews and designs **AI agent systems**. For ordinary source code use the engineering review; for a binding go/no-go verdict use the archetype review (see docs/domains.md → "review appears in three domains").
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.2 KB |
| refs/audit.md | markdown | 4.5 KB |
| refs/design.md | markdown | 4.0 KB |
| refs/frameworks.md | markdown | 3.7 KB |
| refs/review.md | markdown | 4.7 KB |
refs/review.md
View on GitHub# agentic:review — Full Agentic Codebase Review Rubric
Full rubric sourced from the `wicked-garden-agentic-architect`, `wicked-garden-agentic-safety-reviewer`, and `wicked-garden-agentic-performance-analyst` fork skills (former `agents/agentic/*.md`), and `skills/agentic/review-methodology/`. Apply to the target path.
## Step 1: Framework + Topology Detection
Run the detection scripts directly:
```bash
wicked-garden run \
scripts/_run.py \
scripts/agentic/detect_framework.py --path "$TARGET_PATH"
wicked-garden run \
scripts/_run.py \
scripts/agentic/analyze_agents.py \
--path "$TARGET_PATH" --framework "${FRAMEWORK_OVERRIDE:-$DETECTED_FRAMEWORK}"
```
If `--quick` is set, stop here and return the structural summary.
## Step 2: Architecture Assessment (Five-Layer)
Using `skills/agentic/agentic-patterns/` catalog, score each layer PASS / FAIL:
| Layer | Check |
|-------|-------|
| Cognition | Single-responsibility, clear reasoning pattern, prompt versioning |
| Context | Memory scoping, context limits respected, checkpointing |
| Interaction | Tool interfaces defined, rate limits handled, graceful degradation |
| Runtime | Orchestration strategy explicit, error handling, observability |
| Governance | Input/output validation, HITL gates positioned, audit logging |
**Topology health checks** (from `analyze_agents.py` output):
- Circular dependencies → CRITICAL
- God agents (>5 responsibilities) → HIGH
- Orphaned agents → MEDIUM
- Deep nesting (>4 levels) → MEDIUM
## Step 3: Safety Assessment (8-Layer)
Apply the full 8-layer rubric from `skills/agentic/refs/audit.md`.
Focus: tool risk, HITL, PII, prompt injection, failure modes.
Build risk matrix:
| Category | Findings | Critical | High | Medium |
|----------|----------|----------|------|--------|
| Tool Risk | | | | |
| HITL Gates | | | | |
| PII Handling | | | | |
| Prompt Injection | | | | |
| AuthN/AuthZ | | | | |
| Rate Limits | | | | |
| Observability | | | | |
| Failure Modes | | | | |
## Step 4: Performance Assessment
Using `skills/agentic/context-engineering/` and pattern catalog:
**Token analysis**:
- System prompt size and cacheability
- Context window utilization (target: leave 35%+ buffer)
- Opportunities for prompt compression
**Latency analysis**:
```bash
# Find sequential agent calls that could parallelize
grep -r "await.*agent\|agent\.run" --include="*.py" "$TARGET_PATH" -A 3
```
| Metric | Target | Finding |
|--------|--------|---------|
| Avg latency p50 | < 5s | |
| Sequential ops that can parallelize | 0 | |
| Cache hit rate | > 60% | |
| Cost per request | {baseline} | |
**Parallelization opportunities**: list agents that have no mutual dependencies
and can run with `asyncio.gather()`.
## Step 5: Pattern Scoring + Issue Taxonomy
Run the pattern scorer to get quantitative data:
```bash
wicked-garden run \
scripts/_run.py \
scripts/agentic/pattern_scorer.py \
--agents "$AGENTS_FILE" --framework "$DETECTED_FRAMEWORK"
wicked-garden run \
scripts/_run.py \
scripts/agentic/issue_taxonomy.py \
--findings "$FINDINGS_FILE" --agents "$AGENTS_FILE" \
--framework "$FRAMEWORK_FILE" --format markdown
```
**Anti-pattern detection** (from `skills/agentic/agentic-patterns/refs/anti-patterns-design.md`):
| Anti-Pattern | Severity | Detection |
|--------------|----------|-----------|
| God Agent | HIGH | > 5 responsibilities |
| Circular Dependency | CRITICAL | A → B → A in call graph |
| State Mutation | HIGH | Shared mutable context |
| Error Swallowing | MEDIUM | bare `except: pass` |
| Tight Coupling | MEDIUM | Direct class instantiation |
| Duplicated Orchestration | LOW | Same pipeline in 2+ places |
## Output Format
```markdown
## Agentic Review: {Project}
**Date**: {date} | **Framework**: {fw} v{version} | **Path**: {path}
### Executive Summary
{2-3 sentences — overall health, top 2 risks, main recommendation}
### Five-Layer Architecture Scores
| Layer | Score | Top Finding |
|-------|-------|-------------|
| Cognition | PASS/FAIL | |
| Context | PASS/FAIL | |
| Interaction | PASS/FAIL | |
| Runtime | PASS/FAIL | |
| Governance | PASS/FAIL | |
### Safety Risk Matrix
{table from Step 3}
### Performance Summary
{token / latency / cost findings, parallelization opportunities}
### Issue Inventory by Severity
**CRITICAL** (fix immediately):
- {issue} — {file:line} — Fix: {action}
**HIGH** (before production):
- {issue} — {file:line} — Fix: {action}
**MEDIUM** (next sprint):
- …
### 4-Phase Remediation Roadmap
| Phase | Window | Items |
|-------|--------|-------|
| P0 | Immediate | CRITICAL items |
| P1 | This sprint | HIGH items |
| P2 | Next sprint | MEDIUM items |
| P3 | Backlog | LOW + tech debt |
```
Write to `$OUTPUT_FILE` when `--output` is set; otherwise return inline.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |