name: second-opinion description: "Get a dissenting second opinion from a DIFFERENT advisor — one likely to disagree with the first. You name the first advisor (or it uses your last /advice consult); the skill then suggests, or lets you pick, a contrasting expert and delivers their opposing view on the same decision, grounded in their public work. Use after /advice when an answer felt too clean and you want a genuinely different mind to challenge it." argument-hint: "[first-advisor] [topic] — optionally name the second advisor too" model: opus allowed-tools:
- Read
- Glob
- Grep
- Bash
- AskUserQuestion
- WebSearch
- WebFetch
- Skill
/second-opinion — A dissenting view from a different advisor
You got one advisor's take. This brings in a different advisor — deliberately chosen because they'd likely disagree — to give the opposing view on the same decision. The clash is the point: it surfaces the assumptions the first take depended on.
These are simulations in the style of real people, never the real people, and never with fabricated quotes. See $REFERENCE_DIR/voice-guide.md.
Paths
ADVISORS_HOME="${ADVISORS_HOME:-$HOME/.advisors}"
ADVISORS_DIR="$ADVISORS_HOME/advisors"
REFERENCE_DIR="$ADVISORS_HOME/reference"
STATE_DIR="$ADVISORS_HOME/state"
PROFILE="$ADVISORS_HOME/profile.md" # durable context about the person
BRIEF="$STATE_DIR/current-brief.md"
Context gate
Read $PROFILE if it exists — it tells you who you're advising (role, arena, what they optimize for); tailor the dissent to them. Then read $BRIEF if present. If the brief is missing or stale for this topic, offer to run /frame first (invoke frame via Skill). The user may skip — note the thin context if so.
How it works
1. Resolve advisor #1 and the topic
- Advisor #1 (the view to be challenged): from
$ARGUMENTS, else the advisor from the last/adviceconsult in this conversation, else the brief — otherwise ask viaAskUserQuestion. Fuzzy-match name → slug in$ADVISORS_DIR. - Topic: the decision and #1's conclusion — from
$ARGUMENTS, the conversation, or the brief'sleaning. If #1's actual conclusion isn't on record, read #1's dossier and characterize the position they'd most likely take, and say you're doing so.
2. Choose advisor #2 (the dissenter)
- Read the frontmatter (
name,role,domains) and the## Summaryof every$ADVISORS_DIR/*.md— just those (the Summary is the selection digest; fall back to frontmatter if absent). Full dossiers are loaded only for the advisor actually chosen. - Identify the advisor whose worldview most opposes advisor #1 on this specific question — different lens, different values, different track record — not just anyone.
- Suggest that advisor with a one-line reason for the expected disagreement, plus 1-2 alternative dissenters, and confirm via
AskUserQuestion(the user can pick or name their own). If the user already named a second advisor, skip the suggestion and use theirs. If the chosen advisor has no dossier, offer/advice add.
3. Deliver the dissent
- Load advisor #2's full dossier. Answer in their first-person voice.
- Open by engaging advisor #1's position directly: where #2 thinks it's wrong, and why — rooted in #2's own beliefs, frameworks, and track record.
- Make the strongest opposing case (steelman, not strawman). Name the scenario in which following advisor #1 would be a mistake, and the assumption #1's view quietly depends on.
- Stay honest: ground in the dossier, mark extrapolation, no fabricated quotes. If advisor #2 would actually largely agree with #1, say so plainly and offer a better-matched dissenter instead of manufacturing conflict.
4. Close
- The single strongest reason advisor #1 might still be right.
- The one piece of evidence that would settle the disagreement.
- Optionally suggest
/boardroomif the user wants more voices in the room.
Open with: *Second opinion — <Advisor #2>, in the style of their public work, dissenting from <Advisor #1>. Not the real people.*
Hard rules
- A different advisor, genuinely chosen to disagree — not the same person, not a random pick.
- Steelman the dissent. The opposing case must be the strongest one, argued in good faith.
- No fabricated quotes; ground each voice in its dossier; flag extrapolation.
- Don't manufacture conflict. If there's no real disagreement, say so and re-pick.
- No defamation — critiques stay real and sourced.
Task Prioritizer
Productivity
Prioritizes your tasks using Eisenhower, ICE, and RICE frameworks.
Weekly Status Report Generator
Productivity
Generate structured and concise weekly status reports.
Daily Standup Report
Productivity
Generates structured and concise daily standup reports.