Track Specification: GenCpp Dogfood Feedback Loop

Overview

Establish a bidirectional feedback loop where Manual Slop is used to develop gencpp while simultaneously identifying and fixing issues in Manual Slop itself. This track focuses on validating the gencpp project setup and ensuring the feedback mechanism is functional for iterative improvement.

Current State Audit (as of 2026-05-10)

Already Implemented (DO NOT re-implement)

gencpp_project_init_20260510 track configured manual_slop.toml template with [conductor].dir = .manual_slop/conductor
src/paths.py:get_conductor_dir() (line 144-156) supports project-specific conductor directory via TOML override
src/paths.py:_get_project_conductor_dir_from_toml() (line 128-142) parses project TOML for conductor path

Critical Finding: Access Restriction

MCP tools enforce a strict allowed-paths whitelist (see src/mcp_client.py:88-90):

_allowed_paths: set[Path] = set()

The whitelist is populated by configure() from project file_items. By default, only files within C:\projects\manual_slop are accessible via MCP tools.

To access C:/projects/gencpp, one of these must be true:

C:/projects/gencpp is added to the file_items of a Manual Slop session (GUI aggregation), OR
The user manually copies issues/findings into the Manual Slop workspace

Gaps to Fill (This Track's Scope)

Gap 1: No verification that gencpp project directory is accessible and TOML loading works
Gap 2: No feedback loop mechanism for logging issues encountered while using Manual Slop with gencpp
Gap 3: No systematic approach to tracking "found issues" that need fixing in Manual Slop during dogfooding
Gap 4: User cannot directly browse C:/projects/gencpp via MCP tools due to path restrictions

Goals

Verify Manual Slop can load and target the gencpp project at C:/projects/gencpp
Establish a "GenCpp Findings" log mechanism to capture issues found during dogfooding
Create a simple workflow for the user to report issues and have them become track tasks
Resolve MCP path access restriction so the user can browse gencpp via Manual Slop tools

Functional Requirements

Manual Slop must be able to select C:/projects/gencpp as the active project directory
The .ai/ directory configuration must be recognized and loaded properly
A findings/journal mechanism must exist to log feedback during the dogfood session
Issues found should be trackable and actionable
User must be able to browse C:/projects/gencpp via MCP tools without getting "ACCESS DENIED" errors
Alternative: If direct access isn't feasible, provide clear workaround (e.g., copy files to Manual Slop workspace, or use file path manually in prompts)

Non-Functional Requirements

Minimal overhead - the feedback loop should not slow down the coding workflow
All feedback must be captured in a structured format for easy review
Path resolution must work correctly for the C:/projects/gencpp directory

Architecture Reference

src/paths.py - Path resolution with project-specific overrides
src/project_manager.py - Project loading and configuration
docs/guide_architecture.md - Threading model and path resolution

Out of Scope

Fixing specific issues found in Manual Slop (those become separate tracks)
Modifying the gencpp codebase directly (that's the user-controlled target)
Implementing complex automation - this is about establishing the foundation

3.4 KiB Raw Blame History