File Equivalence vs. the Visa Harness

One-for-one comparison of every doc/HTML file, and the honest list of genuine gaps vs. this project.

File Equivalence vs. the Visa Harness Project

A side-by-side check of every markdown and HTML file in the Visa harness project (Visa-Vulnerablity Agentic Harness + its vault) against this project, to confirm parity where it matters and call out genuine structural gaps rather than silently building everything Visa has.

Totals

Visa Ours
Markdown files (excluding Visa's dropped-in source-copy sample) 59 41
HTML files 13 19

We have more HTML because every doc here — not just HLD/LLD/two module docs — was converted to a styled HTML sibling (see task that produced docs/00_START_HERE.html, docs/01_overview.html, docs/design/*.html, docs/tools/graph-intelligence.html, and all five docs/modules/*.html). Visa converted only HLD, LLD, backends, and remediation_agent.

One-for-one equivalents (confirmed present in both)

Visa Ours
architecture.html architecture.html
docs/00_START_HERE.md (+ now .html) docs/00_START_HERE.md + .html
docs/01_overview.md (+ now .html) docs/01_overview.md + .html
docs/design/HLD.md + .html docs/design/HLD.md + .html
docs/design/LLD.md + .html docs/design/LLD.md + .html
docs/design/data-flow.md (+ now .html) docs/design/data-flow.md + .html
docs/modules/README.md (+ now .html) docs/modules/README.md + .html
docs/modules/*.md (per-module docs) docs/modules/*.md (per-component docs) + .html for all five
docs/tools/graph-intelligence.md (+ now .html) docs/tools/graph-intelligence.md + .html
docs/process/* (process + runbooks, inside docs/) samples/process/* + samples/runbooks/* — same content shape, different location by explicit instruction (called "samples", not docs/process/)
scratchpad/modtables/*.md (per-module file/LOC/docstring inventory) scratchpad/modtables/*.md (per-component inventory — Python ast walk for the three Python components, heading/word-count pass for the two prompt-only skills)
tool-outputs/README.md, tool-outputs/codeviz/*.html, tool-outputs/graphify/* same shape: tool-outputs/README.md, tool-outputs/codeviz/*.html, tool-outputs/graphify/*
samples/README.md (Visa's is a dropped-in copy of its own source tree) samples/README.md (ours holds authored process/runbook docs, not a source-tree copy) — same filename, different purpose per the explicit request that created it
diagrams/module-graph.html diagrams/module-graph.html
Vault: MOC, Overview, Architecture, per-topic notes, Project Log Vault: Map of Content, Overview, Architecture (HLD), Closed Loop, Hunt Phases, Fix and Verify, Runtimes, Glossary, Graph Intelligence Findings, Project Log, plus one VulnHunter doesn't have an analog for on Visa's side: Model Selection and Backend (VulnHunter's no-DeepAgents / per-component model-routing writeup)

Genuine gaps (present in Visa, not built here)

Not gaps (same content, different location/name by explicit instruction)

Net read

Structural coverage is at parity or ahead everywhere it affects a reader (HLD/LLD/data-flow/module docs/scratchpad/tool-outputs/vault all have direct equivalents, and HTML coverage is broader here). The one real gap — a portable templates/ library — is a "build if this becomes a template for another project" item, not a missing deliverable for VulnHunter onboarding itself.


See also: Start Here · Module index.