Case study: vault memory, GitHub, and shipping audit

This page summarizes a multi-turn Cursor session: memory_ingest / memory_recall into an Obsidian vault, GitHub triage and new issues, backlog ordering, and implementation of the first enterprise vertical slice — optional audit (#89) — plus docs, Helm, and website alignment.

Context

• ClawQL MCP exposes optional tools behind CLAWQL_ENABLE_* flags.
• Durable context belongs in memory_ingest / memory_recall; ephemeral scratch belongs in cache() — not the vault.
• The session combined external threads (analysis, vision docs, ecosystem posts), live issue lists, and repo work in one arc.

Vault memory ingests

Representative vault pages (stable titles, append: true where notes already existed):

Theme	Examples
Repo / release analysis	Grok cross-check vs v3.2.3, suggested issue titles, follow-on design (schedule, sandbox, synthetics, cache vs memory).
Product vision	ClawQL-Agent roadmap, enterprise (DORA/DACI), x402 / public gateway, payment discovery, clawql.com gateway notes.
Ecosystem	gRPC transport + Gallery / LinkedIn narrative; third-party model tweets (flagged for verification).
Workflow	cache vs memory_ingest — when to use which.
Execution	Open-issues snapshot, #39 → #69 dedupe, #88–#91 created.

Rule of thumb: ingest decisions, debugging conclusions, and contracts; avoid secrets in vault text.

Issues and housekeeping

Action	Detail
Closed duplicate	#39 in favor of #69 (iac_inspect).
New issues	#88 gateway / x402 / discovery; #89 enterprise epic; #90 synthetics + schedule; #91 Gallery gRPC (upstream issues disabled).
Starter artifact	website/public/.well-known/payments.json + /.well-known/* headers — supports #88 / docs deploy #87.

Prioritization snapshot

Discussed ordering across #76–#78, #68–#71, #87, and the new #88–#91. Chosen implementation target: #89 — design plus one shipped tool before broader metrics / governance.

Work completed: enterprise audit (#89)

1. Design: docs/enterprise-mcp-tools.md.
2. Code: clawql-audit — append / list / clear; CLAWQL_ENABLE_AUDIT; wiring in tools.ts.
3. Tests: dedicated audit tests + optional-flags + stdio listing when the flag is on.
4. Surface area: .env.example, root README, CHANGELOG, Helm chart, K8s / Helm docs, website tool pages, Cursor vault skill.

Test bar: npm test green after the change set.

Outcomes and follow-ups

Outcome	Notes
Vault	Searchable graph for roadmap, payments, gRPC — refresh stale cross-links when recalling.
Tracker	#89 remains an epic until additional enterprise tools land; audit satisfies the first slice.
Next	#76–#78, #88 hardening, #91 Gallery PRs, #90 synthetics.
Site	Confirm /.well-known/payments.json on docs.clawql.com after deploys.

References

• docs/case_studies/vault-memory-github-session-2026-04.md — this case study (canonical)
• docs/case_studies/README.md
• docs/enterprise-mcp-tools.md
• docs/mcp-tools.md
• Issues #88–#91, #89