GitHub repo surface consistency¶

Use this file as the durable spec for the GitHub-facing surfaces of openclaw-mem.

Scope note: maintainer-facing guidance in a public repo; it governs outward language but is not itself end-user product documentation.

Goal¶

When someone encounters the repo through GitHub search, pinned repos, releases, or social previews, they should get the same first impression:

openclaw-mem is about trust-aware context packing for OpenClaw.

Not generic memory storage. Not backend internals first. Not a universal graph-schema claim.

Canonical messaging¶

One-line description¶

Trust-aware context packing for OpenClaw — pack only what the agent should trust.

Expanded line¶

Trust-aware context packing for OpenClaw — pack only what the agent should trust, with provenance, receipts, and local-first recall.

Problem statement¶

Long-running agents do not just forget. They also admit stale, weak, or hostile context into future prompts. openclaw-mem narrows around that admission problem.

Public repo messaging sequence¶

Problem: trustworthy memory behavior under long-running pressure.
Approach: trust-aware context packing with explicit trust tiers + receipts.
Proof: synthetic before/after artifact as the first deep link.
Adoption path: sidecar-first install path, optional mem-engine promotion.

Use this order in README / About / release body intros.

GitHub About / repo description¶

Final description¶

Trust-aware context packing for OpenClaw — pack only what the agent should trust, with provenance, receipts, and local-first recall.

Homepage¶

Keep homepage pointed at docs root: - https://phenomenoner.github.io/openclaw-mem/

Why: - root page is the stable landing surface - proof pages are better as strong links from README/docs, not the repo website URL itself

Pinned repo note¶

Pinned repositories do not expose a separate custom blurb beyond repository description.

So the repo description is effectively pin text.

Current pin-safe line:

Trust-aware context packing for OpenClaw — pack only what the agent should trust, with provenance, receipts, and local-first recall.

Short fallback:

Trust-aware context packing for OpenClaw.

Release tagline¶

Short release tagline¶

Trust-aware context packing for OpenClaw: smaller prompt packs, explicit trust tiers, and receipts for why memory entered the pack.

Slightly longer version¶

openclaw-mem now frames its core value around trust-aware context packing: building smaller, safer prompt packs with visible trust tiers, provenance, and receipts.

Use this line near the top of release notes when the release materially strengthens pack selection/provenance/citations/memory hygiene.

Recommended headline¶

Trust-aware context packing for OpenClaw

Recommended subhead¶

Pack only what the agent should trust

provenance
receipts
local-first recall

What the image should imply visually¶

before: noisy / larger / mixed-trust pack
after: smaller / cleaner / trust-gated pack
emphasis on selection, not storage volume

too many flags / CLI snippets
backend nouns as headline (LanceDB, hybrid engine, SQLite sidecar)
giant explanatory walls of text

Suggested alt-text / caption¶

openclaw-mem: trust-aware context packing for OpenClaw — pack only what the agent should trust.

Proof links to pair with GitHub surfaces¶

Use these for second-click depth after the one-line pitch:

Canonical proof: docs/showcase/trust-aware-context-pack-proof.md
Proof metrics: docs/showcase/artifacts/trust-aware-context-pack.metrics.json
Companion demo: docs/showcase/inside-out-demo.md
Copy pack: docs/launch/trust-aware-context-pack-copy-pack.md
Operator checklist: docs/launch/proof-first-relaunch-checklist.md
Release-surface proof pack: docs/launch/release-surface-proof-pack-v0.md

Repository topics guidance¶

Keep topics aligned to positioning first, implementation second.

Preferred set: - openclaw - openclaw-plugin - agent-memory - context-engineering - prompt-packing - provenance - local-first - memory-systems - llmops - observability - sqlite

Backend-specific topics like vector-search, hybrid-search, embeddings, lancedb are true, but should not dominate first impression unless discovery data proves they matter more.

Boundary checks (must stay true)¶

KOL/GTM lanes remain linked-but-separate (no authority merge in this spec).
Query-plane default and action-plane write-gated framing remain explicit.
Graph/reference/knowledge-graph messaging stays a flagship feature family, not an overclaimed universal schema.
For publicly released materials, internal narrative and external narrative must remain clearly separated; do not publish internal framing formulas or backstage positioning language as outward-facing copy.