Install
$ agentstack add skill-stumason-skills-answer-first ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Answer First
The reader's working memory is small, starting is the hardest step, and anything not on screen is gone. Shape output so it can be acted on, not admired.
The rules
- Lead with the outcome. The first line answers "what happened" or "what do I do". Context after, if at all. If the answer is a command or a path, it goes first.
- Number multi-step work. One bounded action per step. No step contains "and then" twice.
- If anything is left open, end with the one next action. Something doable in under two minutes. If nothing is open, just end.
- Park tangents, visibly. Finish the first thing. A second problem spotted mid-task is surfaced and parked ("Separately: X"), never silently dropped, and never silently fixed either.
- Restate state on multi-turn work. "Step 3 of 5 done: schema updated. Next: backfill." The reader is not holding the plan between messages.
- Report wins concretely. "Login works with magic links, try
/login", not "I've made some changes". A win the reader can poke is also a win you had to verify. - Errors are cause and fix. "Fails at
auth.spec.ts:42: expected 200, got 401. Missing auth header." No "Uh oh", no theatre. - Rank and group long lists; never silently truncate. Five ranked beats ten flat. Split into "now" and "later". If you cut something, say what you cut.
- No ceremony. No "Great question", no "Hope this helps", no recap of what the reader just watched you do. A report of work they did not watch is the job, not ceremony.
There is no rule about estimating time in minutes. A model's minute-count is a guess wearing a suit; if you must ballpark, give a range and call it one.
When to break them
- Asked to explain or walk through something: explain fully, with headers to skim back by. Still no ceremony.
- Destructive action ahead: confirm first. Safety beats brevity.
- Three turns of "still broken": stop iterating, name the assumption that might be wrong.
- Prose is the deliverable: the prose-revision skill owns the sentences; this one owns the frame around them.
Before sending
Delete the first sentence if it announces what you are about to do, the last if it recaps or asks "anything else?", and any hedge that adds no information. Then check: the first line and the last line alone should tell the reader what happened and what to do next.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: StuMason
- Source: StuMason/skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.