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 commands/runnithan/claudex-setup/top-tipsgit clone --depth 1 https://github.com/runnithan/claudex-setupWhat 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 | $0.00045 | $0.00811 |
| Opus 5 | $0.00023 | $0.00405 |
| Sonnet 5 | $0.00009 | $0.00162 |
| Haiku 4.5 | $0.00005 | $0.00081 |
Grade A, and why
top-tips 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 — 36 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Top Tips
/optimise delivers lessons that touch a tracked project's config, and it does that well. This command delivers the class that structurally leaks past it (measured 2026-08-30 at a steady ~25% of high-value general lessons): owner-performed, cross-project workflow habits with no config file to patch, which no process revisits once the extraction run that filed them ends. It ranks the corpus's unadopted general lessons and walks the owner through the top N.
Arguments
$ARGUMENTS: N, how many tips to offer this run (default 3).
Ledger
projects/top-tips-ledger.md records every offer and outcome: ADOPTED (with the destination file), DECLINED (with the reason), or LATER (with the date). Never re-offer an ADOPTED or DECLINED entry; re-offer LATER entries after 30 days. Create the ledger on first run if missing. The ledger is what makes each dialog final; without it every run re-litigates the same tips.
Procedure
- Build the candidate set. Active lessons (
status: active) whosetool:includesclaude-codeor is absent. Exclude, in order:- lessons already in the ledger (offered);
- lessons cited in
projects/habits.md, anyprojects/*/*/habits.md, anyprojects/*/*/applied-improvements.md, or encoded in the global~/.claude/CLAUDE.md(adopted); - lessons cited in any
projects/*/*/improvements.mdbacklog (queued: that is/optimise's pipeline, do not duplicate it); - superseded, niche, or project-specific lessons (this command's lane is general and owner-facing).
- Trace via subagents, not in the main session. Fan out Explore subagents over the INDEX by category with the grep corpus (
projects/tree, global CLAUDE.md,.claude/commands/); match by slug AND distinctive phrases, since a CLAUDE.md rule encoding a lesson rarely cites its path. Only finalists get read in full. - Rank the untraced remainder by expected value against how the owner actually works (read
projects/habits.mdandprojects/audit-log.mdfor grounding): exposure of the standing practice the lesson protects, frequency of the situation, and the cost of not knowing. A lesson guarding a risky standing practice (unattended runs, auto-merge) outranks a convenience. - Offer the top N, one dialog each: the tip, why it ranks for THIS owner (grounded in their setup, not generic value), and the exact entry text it would become. Outcomes: Adopt (route per
/optimise§5: agent-performable becomes a config change applied now; owner-performed becomes a habits-file entry), Decline, Later. - Ledger every outcome in the same run and report the remaining untraced count, so the owner knows the depth of what is still unranked.
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 · 36 lines · 45 tokens per session scan A 5ae1a2a77c99
top-tips is a command published in the GitHub repository runnithan/claudex-setup (1 stars, last pushed 2d ago), licensed MIT. It adds 45 tokens to every session and 811 once invoked, about $0.0002 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 commands, from other repositories
suggest-automation
반복 패턴 분석하여 자동화 기회 제안 (주기적 실행 권장).
guide
처음 사용자를 위한 인터랙티브 가이드. 약 3분이면 끝납니다.
setup-pm-skills
Onboard a new user — find out what they do, recommend the right bundles & top skills, and set up a project CONTEXT.md so every skill is tailored to them.
loop-forge
Turn a one-line repetitive task into a reusable, self-guarding slash command (loop-forge).
aliases
CLI aliases for frequently used commands.
setup-rtk
Configure RTK and token optimization for Claude Code.