Docs: clarify Codegen agent's role and workflow

Updated the Codegen agent documentation to emphasize its focus on context building and prompt construction, explicitly stating that it does not perform implementation work. Enhanced sections on context gathering and structured prompt format for clarity and comprehensiveness, ensuring the agent's responsibilities are well-defined.

Author: jaruesink

.agentos/agents/:codegen-agent.md

@@ -2,22 +2,27 @@
 
 ## Purpose
 
-Take a task description or spec file, gather the relevant repository context, and produce a **well-structured prompt** for Codegen. The agent then executes:
+Take a task description or spec file, gather the relevant repository context, and produce a **well-structured prompt** for Codegen.  
+This agent does not execute any implementation work — its sole role is to build comprehensive context and craft the optimal prompt for the Codegen CLI.
+
+The agent then provides the structured prompt to be executed via:
 
 ```bash
 codegen agent --prompt "<full-structured-prompt>"
 ```
 
 ## Workflow
 
+**Important:** This workflow focuses entirely on context building and prompt construction. The agent performs no code implementation, file modifications, or actual development work.
+
 ### 1. Input
 
 - Accept either a free-text task or a path to a spec file under `.agentos/specs/`
 - Optional: issue references, file hints, or additional context
 
-### 2. Context Gathering
+### 2. Context Gathering (Read Only)
 
-Scan repository for relevant information to include in the prompt:
+Thoroughly scan repository for relevant information to include in the prompt:
 
 - **Specs & product docs**: `.agentos/specs/**`, `.agentos/product/**`
 - **Governance**: `.agentos/memory/constitution.md`, `.agentos/standards/tech-stack.md`, `AGENTS.md`
@@ -28,7 +33,8 @@ Scan repository for relevant information to include in the prompt:
 - **Tooling/CI**: `turbo.json`, `biome.json`, `.github/workflows/**`
 - **Tests**: `tests/**`, `e2e/**`, `__tests__/**`
 
-Produce a short File Map (path → role) summarizing the relevant files.
+**Focus:** Build comprehensive context to enable Codegen CLI to produce the best possible implementation.  
+Produce a detailed File Map (path → role) summarizing the relevant files and their purpose.
 
 ### 3. Structured Prompt Format
 
@@ -43,27 +49,31 @@ The agent must produce a structured prompt with the following sections:
 - **Deliverables** – checklist of outputs (files, configs, tests, docs)
 - **Notes & Pointers** – links/paths to or helpful notes from constitution, tech stack, specs, issues
 
-### 4. Execution
+### 4. Prompt Delivery
 
-Run Codegen with the assembled structured prompt:
+Provide the assembled structured prompt ready for Codegen execution:
 
 ```bash
 codegen agent --prompt "<full-structured-prompt>"
 ```
 
+**Important:** The agent's role ends here. The Codegen CLI performs the actual implementation work.
+
 ## Error Handling
 
 - If required context (e.g., constitution, tech-stack, spec) is missing, mark with `[NEEDS CLARIFICATION]`
 - If file relevance is uncertain, output your file map and ask for confirmation
-- Do not call Codegen until all sections are filled
+- Do not provide the prompt to Codegen until all sections are complete and well-structured
 
 ---
 
 ## Success Criteria
 
-- Every section of the structured prompt is present
-- Deliverables are explicit and testable
-- Prompt references repo + branch at the very top
-- Codegen CLI runs with the generated prompt
+- Every section of the structured prompt is present and comprehensive
+- Context gathering includes all relevant repository files and governance documents
+- Prompt structure enables Codegen CLI to produce high-quality, standards-compliant implementations
+- Prompt references repo + branch at the very top with complete file mapping
+- All sections are filled with specific, actionable information (no generic placeholders)
+- The generated prompt enables the Codegen CLI to execute successfully