Process Debt

The LaSasso Tech SEO pipeline worked smoothly because previous projects had already codified the workflow. But six months ago that same pipeline took 3x longer because the pattern hadn't been discovered yet. Execution Templates will prevent rediscovery. But the pipeline still requires human execution: keywords → URLs → rewrites → sheets → email draft.

There are three distinct tiers of improvement work: Memory/Documentation (help sessions know what to do — templates, playbooks, checklists), Infrastructure (automate recurring maintenance — token validators, cron jobs, status drift detection), and Tooling (build CLIs that eliminate the human execution step entirely). We've been strong on tier 1. Tier 2 is emerging. Tier 3 hasn't been addressed yet.

One tier-3 tool would prove the pattern: tech-seo-audit.sh jayco.com --upload-sheets --send-email. Single command, six manual sub-steps gone.

Regulator Content Engine success relied on a framework: claim inventory → fact-check → anti-AI rules → self-audit → adversarial review. But new clients won't know which choices are optional versus mandatory. When do you skip fact-checking? When do you escalate to adversarial review? When do you accept client brand voice that isn't SEO-optimal?

These judgment calls aren't in Execution Templates. They're tribal knowledge. Good templates need two layers — steps (what to do in order) and decision trees (when to skip, when to intensify, when to deviate). A checklist without judgment guidance is a checklist someone will follow wrong.

Example: "Skip fact-checking if: opinion piece, internal memo, client already validated. Intensify adversarial review if: competitor claim, medical/safety statement, new product launch." Layer 2 is what separates a template from guidance.

Improvement #21 (blocker taxonomy) deployed March 20 with simple [BLOCKER: reason] labels. But a live session refined it to four categories — DECISION, EXTERNAL, TEMPORAL, EXECUTION — without a formal change request. The refinement happened organically, not during a sleep cycle.

That's a maturity indicator. Good systems don't need redesign — they get refined organically by users who interact with them. Sessions see blocker labels, understand them better, propose refinements, implement them. That's a virtuous cycle. The system is feedback-looping on itself.

Improvement #21 should be upgraded from 🟡 pending to 🟢 confirmed with live refinement. The taxonomy is actively improving session decision-making.

Five memory file types now exist: MEMORY.md (long-term, ~400 lines), daily notes (800 lines each), improvements.md (unbounded growth), reference-notes.md (catch-all, ~300 lines), and dream logs (one per sleep session). No archive strategy for any of them.

By the end of April, improvements.md could have 60+ entries. By mid-year, 100+. Searching "what recent improvements happened?" becomes O(n). reference-notes.md is already a catch-all holding SEO matrices, article templates, and tech specs — many of which should live in Execution Templates or an entity knowledge graph instead.

Lifecycle strategy: quarterly snapshots for improvements.md, migration of technical notes to templates, quarterly dream summaries, and a new improvements-index.md for quick categorized lookup.

If team OpenClaw launches, new agents and humans will encounter multiple memory files, STATUS.md patterns, open-loops dashboards, blocker taxonomy, and Execution Templates — with no explanation for why the system is designed the way it is. If they don't understand the WHY, they'll implement these as cargo cult: following form without understanding function.

Three layers of documentation are needed for team scaling: Procedures (what we do — Execution Templates ✅), Architecture (how we organize it — mostly missing), and Philosophy (why we designed it this way — entirely missing). Philosophy documents answer "what would break if we changed this?" and prevent well-intentioned simplifications that undermine the design.

Example: "Why multiple memory files instead of one? Cognitive load — 100-line files are scannable, 1,000-line files aren't. Domain separation prevents cross-contamination. Archive-ability matters at scale." That's philosophy documentation.

The tool ecosystem now has three layers: Templates (Execution Templates, 6 planned, 2 deployed), Scripts (google-token-validator.sh built but not integrated; entity scripts deployed), CLIs (none yet — Tech SEO Automation CLI is Improvement #26 hypothesis), Playbooks (Mission Control, Content Engine), and Dashboards (Rank Tracker with 3 clients live).

Improvement #20 proposed adding a "Tools & Playbooks" section to SESSION-CHECKPOINT. But it's still in design phase. Every day that passes is another session that doesn't know the Google Token Validator exists, or that the Content Engine methodology is proven and documented.

Elevation: Improvement #20 should move from design to pending. Add "## 🛠️ Tools & Playbooks Available" to SESSION-CHECKPOINT with a table: Template | When to use | Link | Status.