docs: fix agent and plugin counts across documentation

- Update README.md: 29 agents -> 39 agents (29 file-based + 10 role-based)
- Update package.json: 8 plugins -> 9 plugins
- Update CROSS_PLATFORM.md:
  - Clarify 39 total agents (was showing only 22)
  - Remove Phase 9 review loop from agent list (it's inline orchestration)
  - Add missing perf agents (6) and audit-project agents (10)
  - Correct enhance agent model assignments

All documentation now accurately reflects actual implementation.

Author: avifenesh

README.md

@@ -8,7 +8,7 @@
 
 AI models can write code. That's not the hard part anymore. The hard part is everything else—picking what to work on, managing branches, reviewing output, cleaning up artifacts, handling CI, addressing comments, deploying. **awesome-slash automates the entire workflow**, not just the coding.
 
-**9 plugins · 29 agents · 23 skills · 22k lines of lib code · 1,400+ tests · 3 platforms**
+**9 plugins · 39 agents · 23 skills · 22k lines of lib code · 1,400+ tests · 3 platforms**
 
 ---
 

docs/CROSS_PLATFORM.md

@@ -80,16 +80,15 @@ claude --plugin-dir /path/to/awesome-slash/plugins/next-task
 - `/enhance` - Enhancement analyzer suite
 - `/sync-docs` - Documentation sync
 
-### Available Agents (22 Total)
+### Available Agents (39 Total, key agents shown below)
 
-**next-task: Core Workflow (13 agents)**
+**next-task: Core Workflow (12 agents)**
 
 | Agent | Model | Purpose |
 |-------|-------|---------|
 | exploration-agent | opus | Deep codebase analysis |
 | planning-agent | opus | Design implementation plans |
 | implementation-agent | opus | Execute plans with quality code |
-| Phase 9 review loop | sonnet reviewers | Multi-pass review with parallel agents |
 | deslop-work | sonnet | Clean AI slop from changes |
 | test-coverage-checker | sonnet | Validate test coverage |
 | delivery-validator | sonnet | Autonomous delivery validation |
@@ -100,20 +99,39 @@ claude --plugin-dir /path/to/awesome-slash/plugins/next-task
 | ci-fixer | sonnet | Fix CI failures and PR comments |
 | simple-fixer | haiku | Execute pre-defined fixes |
 
+*Note: Phase 9 review loop uses inline orchestration (orchestrate-review skill) to spawn parallel Task agents*
+
 **enhance: Quality Analyzers (9 agents)**
 
 | Agent | Model | Purpose |
 |-------|-------|---------|
 | enhancement-orchestrator | opus | Coordinate all analyzers |
-| plugin-enhancer | opus | Analyze plugin structures |
+| plugin-enhancer | sonnet | Analyze plugin structures |
 | agent-enhancer | opus | Review agent prompts |
-| docs-enhancer | opus | Documentation quality |
-| claudemd-enhancer | opus | Project memory optimization |
+| docs-enhancer | sonnet | Documentation quality |
+| claudemd-enhancer | sonnet | Project memory optimization |
 | prompt-enhancer | opus | General prompt quality |
 | hooks-enhancer | sonnet | Hook frontmatter and safety |
 | skills-enhancer | sonnet | SKILL.md structure and triggers |
 | enhancement-reporter | sonnet | Format unified reports |
 
+**perf: Performance Investigation (6 agents)**
+
+| Agent | Model | Purpose |
+|-------|-------|---------|
+| perf-orchestrator | sonnet | Coordinate perf investigation |
+| perf-theory-gatherer | sonnet | Generate hypotheses |
+| perf-theory-tester | sonnet | Validate hypotheses |
+| perf-code-paths | sonnet | Map hot paths |
+| perf-investigation-logger | sonnet | Log evidence |
+| perf-analyzer | sonnet | Synthesize findings |
+
+**audit-project: Code Review (10 role-based agents)**
+
+Always active: code-quality-reviewer, security-expert, performance-engineer, test-quality-guardian
+
+Conditional: architecture-reviewer, database-specialist, api-designer, frontend-specialist, backend-specialist, devops-reviewer
+
 **drift-detect: Drift Detection (1 agent)**
 
 | Agent | Model | Purpose |

package.json

@@ -1,7 +1,7 @@
 {
   "name": "awesome-slash",
   "version": "3.5.0",
-  "description": "8 specialized plugins for AI workflow automation (drift detection + AST repo maps) - works with Claude Code, OpenCode, and Codex CLI",
+  "description": "9 specialized plugins for AI workflow automation (drift detection + AST repo maps) - works with Claude Code, OpenCode, and Codex CLI",
   "main": "lib/platform/detect-platform.js",
   "type": "commonjs",
   "bin": {