Home_llm wiki memory template

---

type: index up: "WIKI-INDEX"

llm-wiki-memory-template

Wiki for the llm-wiki-memory-template repository: a template that gives any project a persistent, LLM-maintained, cross-linked Markdown wiki as durable memory, plus the agent-specific surfaces to make the pattern usable from Claude Code, Cursor, or other coding assistants.

This wiki is itself an instance of the pattern it documents. The pages here capture how the template works, why each piece exists, what we have learned from real PRs and reviews, and what remains open.

Navigation

• Index: full catalog of all wiki pages
• Log: chronological record of wiki updates
• Schema: conventions and operations reference

Foundation

• Wiki-LLM-Pattern: the underlying idea, compounding wiki vs RAG, three-layer architecture
• Three-Operations: Query, Ingest, Lint as the codified agent interface
• Template-vs-Derived-Projects: template scaffolding vs instantiated project
• Governance: three-tier model (base / variant / derived project), when to fork a variant, rebase discipline as variants proliferate
• Agent-Overlays: Claude Code, Cursor, none; same pattern, different surfaces
• Typed-Edges-and-Frontmatter: frontmatter conventions, edges as interface operations

Mechanisms

• Knowledge-Graph-Pipeline (PR #14): scripts/kg/, rdflib + pyshacl in-process; SPARQL over wiki frontmatter, optional Fuseki
• Verification-Gate: four-criteria pre-commit gate (PR #1)
• Discipline-Gates: universal rationalizations always wrong (PR #1)
• PostToolUse-Hook: advisory write-time reminder
• Log-Entry-Attribution: by: line + one commit per entry (PR #4)
• Sync-Flow: ALWAYS_FILES contract for propagating template updates
• Test-Harness (in review, PR #10): bash test harness + GitHub Actions workflow for structural validation of template changes
• Multi-Agent-Write-Protocol (aspirational): OCC-with-semantic-retry write protocol for wikis with multiple LLM-agent writers; prevents push-level collisions by construction

Lessons learned

• Lesson-Parallel-File-Drift: same content in two files is a footgun
• Lesson-Hook-Type-Matters: prompt-type hooks cannot be advisory
• Lesson-Validation-Methodology: structural + behavioral testing protocol

Open limitations

• Limitation-Prompt-Injection-Residual: the dominant residual risk
• Limitation-Inverse-Edges-Missing: issue #3, no declared edge inverses
• Limitation-Untested-Overlays: Cursor and none not behaviorally validated

Applied examples

• Applied-Example-Microelectronics-Tutor-Demo: wiki-llm as an undergraduate tutor
• Lessons-Learned-From-model_fusion: field report from a long Claude Code session in PaperAnalyticalDeviceND/model_fusion; surfaces 12 implementation gaps against current template HEAD (0 addressed, 3 partial, 9 missing); recursive-bootstrap finding RESOLVED in PR #25
• Analysis-of-model_fusion-Recommendations: assessment of those 12 recommendations; endorses #1, #4, #8, #9(b) as the high-value cluster (now shipping as PRs #28–#31), pushes back on #2, #6, #10, #11

Governance and meta

• Implementation-Status: current real vs aspirational classification for every wiki page
• Wiki-Corrections-Log: chronological record of corrections made to align the wiki with code

Getting Started

This wiki is human-readable on GitHub or any Markdown renderer. Start at the Index to see all pages organized by category, or read the Schema for conventions when adding new pages.