Current truth first. History kept in context.
The first records are current-source authority. Older routes remain available as historical or reusable research and must not be presented as current product truth.
83ab4ba. The 2026-08-29 corpus is preserved for method, market research and reusable context.All generated documents
Use the current synthesis and programme handoff first; traverse older documents only through their historical/reusable label.
Canonical 83ab4ba product inventory, trust boundary, performance findings, donor ruling and smallest safe sequence.
What HALO is, what the programme achieved, decisions, blockers, evidence limits and the smallest next move.
Every page in Camron's app: what's on it, what we found, and the move — keep/transplant/add/fix/build. The join of his code + all research + the repos.
Recovered Convex program: safe preview, containment, deploy/restore, tests, identity, money, migrations, durable jobs, aggregates, search, and completion gates.
Reconciled counts. Where a lane file disagrees, the lane file is wrong.
The merged read across all seven research lanes.
Findings read directly from source by the main agent, not delegated.
Per-domain maturity grades and the build-vs-buy line.
Ten capability lanes of verified open-source candidates, plus the trap list.
Severity-ranked findings with file:line evidence and fixes.
Where prompt-to-app stopped, and the reusable blocks hiding in the code.
Money representation, compliance and audit gaps for a creator agency.
What the Convex ecosystem makes nearly free that is not being used.
What the commercial platforms ship that HALO does not.
Feature-level teardown of the two market leaders.
Three more platforms and the white-label category.
Unnamed players, adjacent verticals, and the trust-accounting finding.
What the market ships, what it does not, and the pricing inversion.
The GitHub code that does this 100x better — survived adversarial review.
The corrected multi-persona architecture, ranked full systems, niche donors, Buzz composition, and test order.
How block/buzz fits HALO, exact authority split, adapter design, verified gaps, and safe rollout.
Whole-product comparison across CRM, ERP, knowledge, portals, chat, and AI-native systems.
First-party evidence for system-class products, creator-agency tools, and bounded capability vendors.
Workflow-first repository discovery, verified shortlist, licensing boundaries, rejects, and coverage accounting.
Multi-persona operator model, recurring pain clusters, workflow vocabulary, and source-quality labels.
Candidates recovered from the 1.36M RepoCard spine and 307k curated repository corpus.
Workflow-first scope, client-selected Buzz anchor, evidence contract, and research boundaries.
The OnlyFans-management operating model, domain boundaries, evidence rules, and campaign success criteria.
Creator acquisition, onboarding, consent, contracts, portal authority, renewals, access, and safe offboarding.
Chatter queues, playbooks, shift handoffs, fan-reference continuity, QA, attribution, and platform boundaries.
Brief-to-version-to-approval-to-publish content operations, provenance, customs, Drive, and AI-assisted production.
Source events, double-entry liability truth, creator statements, reconciliation, payout rails, and legal gates.
Scenario practice, rubrics, QA calibration, coaching, certification, fair incentives, and creator engagement.
Exception-first agency control, authority-safe metrics, operational evidence, decisions, and closure receipts.
HALO and protected Oracle patterns that transfer into the OFM control plane—and the boundaries that do not.
Generator: site/scripts/generate-docs.mjs · source files remain outside site/system-map/.