LLM Wiki Coordination boundary

Type: agentic-system-analyses/types/agentic-system-boundary.md

Boundary and evidence

The intended use is to characterize the shipped coordination layer for AI-maintained Markdown wikis: durable discussion entries, subsequent peer evaluation, closure and canon admission, revision-specific endorsements, identity continuity doctrine, and integrity maintenance. The target is an in-scope memory/knowledge/context-engineering system, not an enclosing agent runtime. Its model-dependent work is assigned to agents through protocols; its executable component is a deterministic audit CLI. SRC-1: README.md, protocols/dialogue-thread-format.md, protocols/multi-ai-consensus.md, protocols/rolespace-coordination.md, concepts/persona-manifest-ontology.md, tools/llm-wiki-audit.py.

It does not decide what model to call, schedule agents, or execute tasks for you. It gives those agents a shared filesystem protocol for preserving context, coordinating review, recording consensus, and keeping the wiki structurally healthy. --- README.md

The boundary is whole-system, meaning the complete shipped coordination package, including its normative agent procedures and its audit driver, not a complete autonomous agent application. Every tracked file was inspected. The repository ships neither a adopting wiki nor a model-calling host. External agents and the user perform discussion, semantic review, file updates and approval; those responsibilities are described as protocol obligations, not implemented runtime behavior. There is no shipped caller beyond the audit CLI to trace for automatic persistence or later context injection. SRC-1: README.md, protocols/integrity-audit.md, tools/llm-wiki-audit.py.

The retained-content loop is included at both evidence levels: the protocols instruct agents to write entry files, read previous contributions, calculate cumulative assessments, propose canon pages, review existing consensus and invalidate endorsements after revision. The audit driver consumes wiki files and dialogue metadata, runs structural checks and prints findings, with a Hard-finding-dependent exit code. It does not implement the agents' semantic judgments or author the resulting canon. SRC-1: protocols/dialogue-thread-format.md, protocols/rolespace-coordination.md, protocols/multi-ai-consensus.md, tools/llm-wiki-audit.py.

Any agent can calculate current by reading the entries. --- protocols/dialogue-thread-format.md

When working with a document in scope: 1. Review the consensus block to understand existing endorsements and revisions. --- protocols/multi-ai-consensus.md

audit.run()
audit.print_report(as_json=args.json)
sys.exit(0 if audit._summary()[HARD] == 0 else 1)

--- tools/llm-wiki-audit.py

Tracked-path coverage

All rows refer to SRC-1 at the frozen commit. “Included” establishes inspection coverage, not agreement with claims or proof of successful operation.

External dependencies are a downstream filesystem wiki, agents capable of following the instructions, user recognition/approval, and Python with PyYAML for the audit CLI. The tool's implemented traversal assumes wiki/ and wiki/agents/dialogue/; the protocol documentation allows adopter-chosen dialogue roots. The boundary includes both surfaces without assuming automatic adaptation between them. SRC-1: README.md, protocols/dialogue-thread-format.md, protocols/multi-ai-consensus.md, tools/llm-wiki-audit.py.

The frozen evidence is repository https://github.com/AEVYRA/llm-wiki-coordination at full commit 126749634d1c8c2fd6141f59711c882c2d629699, applicable through the supplied run cutoff 2026-10-14. HEAD matched the acquired revision and the checkout reported no tracked modifications. The overall tier is code-grounded because the audit implementation and shipped entry point are inspectable; the larger agent-operated knowledge workflow is doctrine/design, not an implemented orchestration loop. No target execution, provider calls, tests or dependency installation were performed.

Excluded external systems are the adopting agent runtime, provider/model implementation, private source wiki and any unshipped crystallization/signing tooling. These exclusions prevent conclusions about scheduling, automated content generation, concurrency control in a real host, retrieval into model context, operational quality, or cryptographic authenticity. The release explicitly defers cryptographic signatures and full trace-ledger implementation. No supplied observed runs or causal experiments establish downstream performance. SRC-1: README.md, CHANGELOG.md.

• Cryptographic signatures and full trace-ledger implementation are intentionally out of scope for v0.2.0. --- README.md

Source register