From lurii-cc
Incremental updates to `memory/` (profile, targets, platforms, patterns, on-horizon). Edits YAML frontmatter + markdown body via `Edit`, appends `CHANGELOG.md`, confirms before destructive edits. Also receives handoffs from `portfolio-advisor`. Use when the user says "update memory", "remember X", "target changed", "I sold/bought Z", "new platform", "mark Y executed", or invokes /memory-update.
How this skill is triggered — by the user, by Claude, or both
Slash command
/lurii-cc:memory-curatorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Maintain the `memory/` folder at the repo root. Each file has two layers: **YAML frontmatter** (canonical for skill consumption) and **markdown body** (human prose). Frontmatter is the source of truth — body may lag. Minimal diffs, explicit confirmations on risky edits, append to `memory/CHANGELOG.md` on every change. Honor the schema in `memory/README.md`.
Maintain the memory/ folder at the repo root. Each file has two layers: YAML frontmatter (canonical for skill consumption) and markdown body (human prose). Frontmatter is the source of truth — body may lag. Minimal diffs, explicit confirmations on risky edits, append to memory/CHANGELOG.md on every change. Honor the schema in memory/README.md.
| File | Scope |
|---|---|
memory/profile.md | Who the investor is: location, goals, horizon, risk, cash flow, portfolio scale, cadence |
memory/targets.md | Allocation targets and deployment priority |
memory/platforms.md | Brokers, exchanges, yield venues, wallets, regional blocks |
memory/patterns.md | Investment principles, MCP data quirks, analysis patterns, style preferences |
memory/on-horizon.md | Active plans and pending actions |
memory/CHANGELOG.md | Dated log of every curator action |
Classify the user's update, then pick the target file:
profile.mdtargets.mdplatforms.mdpatterns.mdon-horizon.mdIf unclear, ask the user which file (or which of several to update).
Edit tool, not Write. Preserve surrounding content.targets.md (weight_pct, range_pct, gap_thresholds.*)patterns.principles[]platforms.blocked[]memory/README.mdFor these, show the exact diff and ask "Confirm?" before applying.
CHANGELOG.md.on-horizon.md items have three lifecycle states:
on-horizon.md; add CHANGELOG entry with date + detailon-horizon.md; CHANGELOG entry with reasonDetect automatic transitions when possible: if the user mentions executing a rec (e.g. "sold " or "deployed "), propose the transition.
Append to memory/CHANGELOG.md. Newest entries at top. Each entry:
## YYYY-MM-DD
- **{file.md} — {field or section}:** {one-line description of change}. {Optional rationale.}
Group multiple edits from a single session under the same date heading.
Examples (illustrative only — concrete values come from the user's actual memory):
## YYYY-MM-DD
- **targets.md — targets[<asset>].weight_pct:** <old>% → <new>% (rationale).
- **on-horizon.md — items[<id>].status:** active → executed; <executed-quantity/value>.
- **platforms.md — yield_venues[<venue>].apy_estimate:** <old>% → <new>% (rate cycle change).
Edit.Edit.patterns.output_language is set, use it for body prose. Tickers, platform names, code identifiers stay in ticker_language.Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub chizhovyui/lurii-cc --plugin lurii-cc