version: "1.0.0" name: why-not description: The discipline, basis, and operations for maintaining the root WHY-NOT.md in the zk-memory repo — the comparison of zk-memory (the library, NOT the hermes-zk-memory wrapper) against other agentic-memory systems. Use when adding or updating a "why not X" entry, when a new memory system is proposed, when a competitor materially changes, or when asked why we don't use some other tool (mem0, LangMem/Letta, LlamaIndex, Khoj, Chroma, prospecta, QMD, Obsidian, etc.). Triggers: why-not, why not X, compare to mem0/LangMem/Khoj/QMD, should we use X, what about X.

why-not — maintaining WHY-NOT.md

WHY-NOT.md is the repo's answer to "why don't you use X?" for zk-memory — the host-agnostic library. It is not about hermes-zk-memory (a thin wrapper; its only job is Hermes plumbing).

The file is the application of `PRINCIPLES.md` to named systems — it exists so the merits (Z-principles, P2/P3/P9) have a place to defend themselves against recurring suggestions, and so the reasoning isn't re-derived from scratch every time someone asks.

Discipline (the contract)

1. Anchored, never free-floating. Every "why not" cites the Z/P-principle(s) it protects

(e.g. Z1 write-time judgment, Z9 shared corpus, Z10 capture-fast-integrate-later, P2 standalone, P3 LLM-as-injected-callable, P9 thin adapter). The doc must not invent reasons that aren't in PRINCIPLES — if you want a reason that isn't there, add/change the principle first. This is what stops the doc from rotting: the anchor can't drift from the design.

1. One decisive reason per system — not a feature matrix. Name what it is in one line, then the

single reason(s) we're not it. No scorecards, no "X has this but we have that" tables.

1. It is about the LIBRARY. Don't compare wrappers/deployments; compare the core capability and

design intent of zk-memory against the other system.

1. Never duplicates PRINCIPLES.md. PRINCIPLES is the anchor; WHY-NOT is its external-facing

application. Two different jobs, one source of truth.

1. Dated. Header carries the date it was last reviewed/updated, matching PRINCIPLES.

Basis (what a good entry looks like)

• What it is — one line, concrete (language/runtime + the one thing it does).
• The decisive mismatch — anchored to a principle. Prefer the strongest, most durable reason,

not a list of minor ones. "It's Node and we're Python" (P2) beats "its CLI flags differ."

• What it'd cost us — the concrete tradeoff (infra, runtime, recall, write-side judgment).
• Keep it current — a competitor changing is a trigger to re-check the entry, not to delete it.

Operations

• Add an entry when a system is proposed (in a review, an issue, a conversation). This skill's

trigger is the queue: "why not X" / "should we use X" → add or update the entry.

• Update an entry when a competitor materially changes (new runtime, dropped feature, pivots).

Re-verify the decisive reason still holds; if it does, refresh the date; if it doesn't, revise.

• The maintenance rule lives in AGENTS.md so future agents keep the file honest: "Why not X? See

WHY-NOT.md — add a line when a system is proposed, update when one changes."

• Linking: AGENTS.md links to WHY-NOT.md; WHY-NOT.md links back to PRINCIPLES.md (the anchor).

References

• WHY-NOT.md — the file this skill maintains.
• PRINCIPLES.md — the anchor every entry cites.
• AGENTS.md — the charter; carries the maintenance rule and the link.