Work / Engineering framework

AI Command Center

Long-running engineering work needs a clear place for permanent rules, current task state and records of what happened.

My implementation: Markdown architecture and role documents, starter workspace templates, six SQLite tables and two Python utilities.

Public framework, templates and supporting utilities; not a deployed agent orchestration service.

Context

Shared AI workspaces can accumulate conflicting instructions, stale handoffs and unclear ownership of files.

Problem

Long-running engineering work needs a clear place for permanent rules, current task state and records of what happened.

What it enables

Provides a concrete starting structure for task ownership, handoffs, documentation freshness and recorded changes.

Constraints

Public framework, templates and supporting utilities; not a deployed agent orchestration service.
Authority and append-only policies are documented conventions, not enforced security boundaries.

System

A documented operating model with workspace templates, a SQLite record schema and small Python utilities for coordinating AI-assisted engineering.

Inputs

Workspace rules and architecture decisions
Current task handoffs
Commit, task and pipeline-run records

Implementation

Separate permanent rules, project decisions and rolling task memory.
Define roles and change classes in Markdown templates.
Provide SQLite tables for records and Python utilities for document-age checks and advisory lock files.
Markdown architecture and role documents, starter workspace templates, six SQLite tables and two Python utilities.

Decisions

Keep permanent rules outside rolling memoryA handoff can expire without discarding the workspace's long-lived constraints.
Use plain files and SQLiteThe structure can be inspected and adapted without requiring a hosted coordination platform.
Document authority separately from evidenceA record of a change describes what happened; it does not itself grant permission for another change.

Artifacts

CREATE TABLE IF NOT EXISTS evidence_correlations (
    correlation_id TEXT PRIMARY KEY,
    source_type TEXT NOT NULL,
    source_id TEXT NOT NULL,
    related_artifact_path TEXT NOT NULL,
    notes TEXT,
    created_at TIMESTAMP NOT NULL
);

Excerpt from templates/.remember/schema.sql. This defines a record structure; append-only behavior is a policy rather than a constraint in this schema.

Evidence

Evidence classes distinguish what can be inspected directly from what is documented, simulated or reconstructed.

README REPORTED
Architecture and operating model
docs/ARCHITECTURE.md
VERIFIED CODE
SQLite record schema
templates/.remember/schema.sql
VERIFIED CODE
Documentation staleness checker
tools/check_doc_staleness.py
VERIFIED CODE
Advisory lock-file utility
tools/file_lock.py

Limitations

The SQLite schema does not enforce immutability through triggers or access controls.
The lock helper checks then writes a file; it is not an atomic mutex or a guarantee against concurrent writes.
Public files do not independently establish organization-wide adoption, productivity gains or production safety.

Discuss a system like this.

Start a work enquiry All work