Agentic Workflows
Formalize and modernize the Claude workflow tooling across the Qpoint project family
Why
A July 2026 audit of every .claude/ directory across the project family found a strong inner loop (subagent design, context engineering, cycles-as-spec) but a weak outer loop (copy-based template sync, no hooks, no automation, inconsistent skill formats, WIP sprawl). Captured as Stage 1 and deliberately parked until Mark could give it full attention.
Objective
Never formally kicked off. The audit and proposed workstreams live in initial-prompt.md.
Scope
See initial-prompt.md — plugin distribution, hooks, scheduled automation, skill-format normalization, staleness signals, opt-in pipeline orchestration, agent philosophy doc, agent evals, permission hygiene, knowledge-layer ownership, brand MCP distribution.
Key Changes
- 2026-08-06 — Stage 1 capture only; parked by design.
Outcome
Superseded by c69.operating-system (2026-08-15), which executed the retro/audit half — skill-format normalization to SKILL.md directories, WIP/staleness cleanup (archive sweep), convention codification (PICKUP.md, thread files), and the methodology literacy layer. The remaining outer-loop workstreams (plugin distribution, hooks, scheduled automation, agent evals, permission hygiene, brand MCP distribution) stay captured in this cycle's initial-prompt.md as source material for future cycles.
Artifacts
- initial-prompt.md
- initial-prompt.md