Dispatch-ready repository queue
Capture: kb/sources/.snapshots/karpathy-gist-comments-2026-11-07. These are analysis candidates, not verified system findings.
All selected sources returned public repository metadata and a substantive README.
No matching current publication/review filename or completion-workshop ownership was found;
this is identity screening, not proof against every possible alias or external run.
Use dispatch-urls.txt or dispatch.json for source inputs. The order is a suggested
first wave (up to 10) followed by a wider queue, not authorization to launch every target.
Do not send selection reasons, inspection questions, old triage or this file to fresh analysts.
Give each worker only its repository URL and the instruction to execute the current
analyse-agentic-system skill with default profile pi-sol, in its own prepared worktree.
No revision is requested: acquisition freezes the default-branch tip at run time.
This screening did not clone repositories, execute target code or freeze a source commit.
| # | Canonical source | Selection rationale (a lead) | First inspection question | Gist evidence |
|---|---|---|---|---|
| 1 | https://github.com/liushengyu/memexos | Review-separated wiki updates and use-feedback maintenance | Which checks and independent review decisions actually gate wiki promotion? | 6452852 |
| 2 | https://github.com/Obelyth/cortex | Commit-bound quote verification and guest proposal boundary | What does quote verification establish, and who can bypass proposal review? | 6365460, 6365468, 6401465 |
| 3 | https://github.com/francisdu53/exia-seed | Graph activation recall and source-line admission checks | How do retained raw evidence, consolidation and automatic recall connect? | 6448248 |
| 4 | https://github.com/suwonleee/llmwiki | Cross-harness capture, decision recall and edit-time rules | Which harness paths deliver decisions into later context and edits? | 6265997, 6284942, 6321505 |
| 5 | https://github.com/alfadur7/llm-wiki-newsroom | Separated author/reviewer roles and proposed rule improvement | Are editorial promotion and rule-change regression gates enforced or prescribed? | 6218432, 6223045 |
| 6 | https://github.com/crow-hq/llm-wiki-v2 | Typed classification before model-authored wiki updates | Which decisions belong to the classifier, and how are mistaken classifications corrected? | 6427891 |
| 7 | https://github.com/nfeldman/amanuensis | Durable code accounts with challenge and freshness state | How do challenged findings become accepted or stale and reach later agents? | 6344820 |
| 8 | https://github.com/demartinogiuseppe/antinomia | Contradictions represented as maintained objects | What closes a tension, and how is resolution evidence retained? | 6207255 |
| 9 | https://github.com/XBlueSky/cortexes | Capture, bounded distillation and fusion into existing memory | Which checks prevent incomplete or duplicate session evidence from replacing memory? | 6284110, 6317008, 6382464 |
| 10 | https://github.com/frankchu91/mindbase-llm-wiki | Human-approved wiki changes and quote-checked lint | Where are source/wiki authority and approval boundaries enforced? | 6328804, 6368378, 6430961 |
| 11 | https://github.com/equationalapplications/curated-thoughts | Shared Markdown/SQLite memory with MCP proposal flow | Which representation is canonical and how do reviewed proposals reach agent reads? | 6144258, 6337308, 6377303 |
| 12 | https://github.com/madgodinc/mgi-mind | Deterministic conflict admission and stale/quarantine states | What inputs authorize replacement, confirmation and quarantine at write time? | 6340323 |
| 13 | https://github.com/sturlese/stigmergy | Team entity stewardship and audience-scoped memory reads | Do admission and visibility checks bind the same bytes that are committed and retrieved? | 6301167, 6306038 |
| 14 | https://github.com/RightL/RightMemory | Structured Git-backed memory and controlled shared views | How are private roots, shared views and multi-agent updates separated? | 6224846 |
| 15 | https://github.com/serradura/okf | Skill, deterministic CLI and maintenance hook split | Which invariants are checked mechanically and which require the host agent? | 6255287 |
| 16 | https://github.com/wikibrain-app/wikibrain | MCP wiki and server-side agent sharing one tool surface | How do instruction reads, pending-source computation and server-side writes interact? | 6362239 |
| 17 | https://github.com/sidleo/llm-wiki | One core across Pi, plugin and CLI with directory write rules | Do all hosts apply the same verification and write authority rules? | 6371101 |
| 18 | https://github.com/nimbalyst/nimbalyst-wiki | Operator-facing decision and open-question memory | How are significant decisions selected, attributed and restored into continuing work? | 6436773 |
| 19 | https://github.com/jimmybackend/MCMA-OpenMemory | Portable memory separate from generated wiki views | How are validation, provenance and encryption preserved across storage adapters? | 6348262 |
| 20 | https://github.com/azaylamba/repocodex | Code-state-bound rationale and freshness attestation | What does a pin check prove and what triggers rationale revision? | 6366954 |
| 21 | https://github.com/OrangeproAI/orangepro-mcp | Maintained behavioral knowledge and change verification | How does retained behavioral knowledge affect later verification and get corrected? | 6237886, 6263632, 6295034 |
| 22 | https://github.com/JanYork/llm-wiki-cli | SQLite-canonical memory graph with Markdown projection | How are agent writes, revisions and citations kept coherent across projections? | 6288782, 6289082, 6308835 |
| 23 | https://github.com/ksx-cz/khms-memory | Immutable typed cards and hook-pushed recall | How do supersession and review decisions change what later sessions receive? | 6338688 |
| 24 | https://github.com/alphazede/bran | Deterministic knowledge traversal and bounded context packets | How does the maintained schema affect ranking and host-agent context delivery? | 6302004 |
| 25 | https://github.com/mohammadmaso/kherad | Git-backed wiki compilation through merge requests | Which review and merge routes distinguish generated drafts from canonical pages? | 6262844 |
| 26 | https://github.com/sturlese/hippocampus | Small injected cache plus index-guided retrieval | What refreshes the cache and what happens when index-guided retrieval fails? | 6264580 |
| 27 | https://github.com/Sistema2D/FrameCode-VibeWork | Carryover executable-spec governance and technical memory | Which activation artifacts deliver its governance into target workspaces? | 6152026, 6155060, 6186108 |
| 28 | https://github.com/giodra96/project-wiki | Code/spec/ADR traceability and progressive context routing | What detects stale code links and how does routing deliver relevant constraints? | 6321649, 6381463 |
| 29 | https://github.com/equationalapplications/expo-llm-wiki | Portable SQLite memory engine with configurable ontologies | Which schema constraints bind extraction, persistence and later graph reads? | 6148265, 6211662, 6308450 |
| 30 | https://github.com/iwe-org/iwe | Structural Markdown graph operations and validation | How do structural editing and schema checks constrain agent-maintained knowledge? | 6317613 |
Source identity notes
serradura/okf-gemredirects toserradura/okf.kostey/khms-memoryredirects toksx-cz/khms-memory.
Recheck active writer/run ownership and the committed analysis method before dispatch. Consult the workshop framing for cross-target scope notes and preparation gates.