Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/ramboz/jig/memory-syncnpx skills add ramboz/jig --skill memory-syncgit clone --depth 1 https://github.com/ramboz/jigWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00228 | $0.03367 |
| Opus 5 | $0.00114 | $0.01684 |
| Sonnet 5 | $0.00046 | $0.00673 |
| Haiku 4.5 | $0.00023 | $0.00337 |
Grade A, and why
memory-sync scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 246 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Spec 002 (memory layer) is fully closed — all four slices DONE: 002-01 (explicit-sync), 002-02 (lookup-pattern), 002-03 (auto-detect-hooks), 002-04 (reconciliation-integration). 002-04's reconciliation integration is now the Memory-sync gate in the spec-workflow reconciliation checklist.
What this skill does
Persists session-derived context to the memory layer via a deterministic helper.
Claude makes the what / where decisions; memory.py does the file I/O,
idempotency, and self-healing of missing memory structure.
When to invoke
- User says "remember this", "save this for later", "add this to the glossary", "note this down", or similar (→ persist flow below).
- User explicitly invokes
/jig:memory-sync. - An unknown capitalized reference appears in the conversation (→ lookup-pattern flow below).
- Session-end consolidation (after slice 002-03 auto-trigger ships).
- The session settled a non-spec shipped decision — a UI string, visual/CSS choice, translation correction, or scoped brand/icon call made outside a spec slice (→ lightweight-decision flow below). This is the forcing function for out-of-spec work, which has no reconciliation phase to catch it.
Lookup-pattern flow
When you see a capitalized reference, acronym, or project-specific term you don't recognize, follow this flow before asking the user:
seen unknown reference X
↓
python3 memory.py lookup "X" .
↓ exit 0 → use the printed definition; do not ask
↓ exit 2 → ask the user once: "I don't recognize X — what is it?"
↓ user answers
↓
python3 memory.py add-term "X" "<definition>" . (or promote if high-frequency)
↓ next time X appears, lookup hits
Concretely, the commands are:
python3 "${CLAUDE_PLUGIN_ROOT}/skills/memory-sync/memory.py" lookup "<term>" "<target>"
# exit 0 = hit (definition + source on stdout)
# exit 2 = miss (proceed to ask the user)
The lookup is case-insensitive and checks hot cache first, then glossary. Hot cache hits win when a term exists in both (the user has explicitly elevated it).
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 246 lines · 228 tokens per session scan A e10cf6dca870
memory-sync is a skill published in the GitHub repository ramboz/jig (6 stars, last pushed 2d ago), licensed MIT. It adds 228 tokens to every session and 3,367 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
frappe-lms
Frappe LMS customization guidance for courses, batches, lessons, quizzes, assessments, enrollment, progress, certificates, portals, and learning workflows. Use when work touches Frappe LMS or education/training flows.
baoyu-comic
Knowledge comic creator supporting multiple art styles and tones. Creates original educational comics with detailed panel layouts and batch-capable image generation. Use when user asks to create "知识漫画", "教育漫画", "biography comic", "tutorial comic", or "Logicomix-style comic".
learn
Knowledge compiler. Extracts patterns, decisions, and anti-patterns from completed campaigns and evolve cycles, then compiles them into structured wiki pages that integrate with existing knowledge rather than appending isolated files. Implements flush→compile→lint pipeline. Auto-triggered by /postmortem and /evolve…
zapier-demo
Walk a new user through setting up their first Zapier action and running it live — the smallest possible win. Asks what app they use, recommends one read action to enable, guides them to mcp.zapier.com to add it, then demonstrates it working in the same chat. Use when the user asks "show me how Zapier works", "set up…
logical-thinking
Determine what follows from explicit premises, rules, evidence, documentation, or observations by selecting and executing the appropriate inference structure. Use when validity, entailment, contradiction, causality, probability, explanation, or the exact strength of a conclusion is central. Do not invoke merely to…
learning-coach
中文高级私人学习教练。用户说“开始学习 X”“继续/下一课”“学完了”“记录进度/打卡”“我学到哪了/还要多久”，正在通过真实项目、源码、作品、题目或任务学习编程、Agent、写作、语言、考试、研究或职业技能，或对具体概念和材料连续追问时使用。触发后主动读取学习现场、学习记录、进度、用户背景和真实材料，替用户选择最近发展区与入口，一次只给一个动作或问题，并用解释、实践、反馈和迁移证据推进掌握。不要让用户先设计课程、选择材料或填写长问卷；只有方向取舍本身尚未解决且会造成明显浪费时，才交给 focus-coach。.