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/lfyxhappy/lfcode/asknpx skills add lfyxhappy/lfcode --skill askgit clone --depth 1 https://github.com/lfyxhappy/lfcodeWhat 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 | $0.00054 | $0.00863 |
| Opus 5 | $0.00027 | $0.00432 |
| Sonnet 5 | $0.00011 | $0.00173 |
| Haiku 4.5 | $0.00005 | $0.00086 |
Grade A, and why
compose:ask 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 yesterday.
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.
This is a copy
100% identical to compose:ask — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Asking the User
The Rule
Every time you need the user to decide, clarify, or approve something, route it through the question tool. Never stop the loop with a natural-language question ("Does this look right?", "Should I proceed?", "Which would you prefer?"). A natural-language question ends your turn without finishing the task; a question tool call does not.
This means: the loop only ends when the task is actually complete — never because you paused to ask in prose.
How to Ask
- Structured options — when the decision has known choices, list them as
options(each with a shortlabeland adescription). - Open-ended — when you can't enumerate good options, pass empty
options. An empty options list renders as a free-text prompt: the user types whatever they want. So anything you'd normally ask in prose can be asked throughquestioninstead. - One question per concern — don't bundle unrelated decisions; ask them as separate questions (or separate calls).
- Don't repeat the question in prose — the tool already renders it. Just call the tool.
question({
questions: [{
question: "Which auth strategy should I use?",
header: "Auth",
options: [
{ label: "Session cookies", description: "Server-side sessions, simplest" },
{ label: "JWT", description: "Stateless, good for multiple services" },
],
}],
})
When No User Is Available
There are two situations where you won't get a human answer. The decision behavior is identical in both — you pick the best option for unattended/headless execution yourself and keep going. They differ only in whether the question tool is reached this turn:
- Question tool absent (e.g.
run/eval, wherequestionis denied) — the tool isn't in your list. You never call it; decide and proceed directly. [Never-Ask]response (never-ask is on) — you do callquestion, but instead of a user answer the tool returns a[Never-Ask]directive. Re-pick from the options you proposed, explicitly state your choice and reasoning in your response text, and continue.
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.
- yesterday First seen · 59 lines · 54 tokens per session scan A 3bb9d1e20600
compose:ask is a skill published in the GitHub repository lfyxhappy/lfcode (2 stars, last pushed 1mo ago), licensed MIT. It adds 54 tokens to every session and 863 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to compose:ask, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
Review a GitHub PR (via gh)
Review a specific GitHub pull request with gh — fetch the diff, fan out reviewers, consolidate, and optionally post the review. Requires the gh CLI or the GitHub MCP server.
Implement (multi-agent loop)
Orchestrate an implement -> review -> fix loop with subagents until reviewers sign off.
Best of N (parallel attempts)
Delegate N parallel subagents on the same task, then pick the best result.
Check work (verify against criteria)
Verify an implementation against acceptance criteria with a reviewer and a tester.
Commit (clean, conventional)
Stage the right changes and write a clear, conventional commit message.
Design doc (write -> review loop)
Draft a design document and iterate writer/reviewer subagents until consensus.