matrix

Price this on the value/effort matrix and commit to a quadrant: $ARGUMENTS

First — measure, don't estimate from vibes

Go read the thing before you price it. An effort number you invented is worse than no number, because I'll act on it. Spend the tool calls:

• Named a ticket? Read it and its comments. The proposal in the summary and the reality in the comments often disagree, and the comments are usually righter.
• Named a doc or a playbook? Read it, then check whether it's still true. A parked plan rots: the fix may be half-done, the blocker may be gone, the tool may have changed. Say so explicitly if the doc is stale or mistaken — that finding is often worth more than the pricing.
• Named work in a repo? Look at what it actually touches. Count the files, the repos, the deploys, the approvals. "One line in many repos" is a different animal from "one line in one repo."
• Is there a cheaper variant nobody costed? Price that too — it frequently wins.

If the project has a record-reconstruction skill (one that pulls ticket, chat and git history on a subject), chain it rather than re-deriving the history yourself.

The two axes, in the user's actual currency

Effort is not story points. It's:

• Hours, and whose: if only one person can do it, their time is the real constraint.
• Prod deploys and approvals required — these cost calendar time and risk, not minutes.
• Blast radius if it goes wrong, and how reversible it is.
• Whether it needs someone else (an approval, a credential, a vendor, someone unavailable). Work that depends on an absent human is high-effort even when the diff is tiny.

Value is:

• Recurrence — does this stop a thing that keeps happening? A 90-minute fix that ends a recurring outage beats a 20-minute fix that tidies one file.
• What breaks, and for whom, if it's never done. Name the blast radius concretely.
• Whether it unblocks other work or is terminal.
• Whether the pain is already being absorbed by a workaround that works. A workaround that holds is real value already banked — it lowers the value of automating it.

Count cost of delay separately when it's real: something that gets more expensive the longer it waits (an expiring credential, an approaching deadline, the one person who knows it becoming unavailable) deserves saying out loud, because it can promote a low-value item on timing alone.

Output shape

Verdict — one bold line: the quadrant and the call.

DO NOW — high value, ~2 hours, and it ends a recurring system-wide outage.

Value: <high/medium/low> — 2–4 bullets, each concrete. What breaks without it, how often, for whom. Cite the real numbers you found.

Effort: <high/medium/low> — 2–4 bullets. Hours, repos touched, deploys and approvals needed, what could go wrong. Name the single most expensive part.

Quadrant — one of:

A batch triage of side findings runs this same grid (do now / fill-in → fold into the open diff, plan it → file a ticket, don't → drop), with two deliberate differences: it measures effort against the diff already open, and only critical severity earns a ticket there. So a PLAN IT here may be a "log it, don't ticket it" in triage. That's the batch budget working, not a contradiction.

Confidence — one line separating what you measured from what you estimated. Effort is almost always the softer number; say which parts you'd bet on and which could double.

Cheaper variant — only if one exists. The 20%-effort version that captures most of the value, priced the same way. If the cheap variant wins, say that in the verdict.

Then — one closing affordance: /go to do it now, or /ticket to file it, or, for a DON'T, name what would have to change to make it worth revisiting.

Rules

• Commit to a quadrant. "It depends" is not an answer. If it genuinely straddles a boundary, pick the side you'd act on and name the one fact that would flip it.
• Never rank it against the board. Emit the placement and the evidence; whoever owns the backlog owns the ordering. "This is more important than ticket X" is not yours to say.
• File nothing, change nothing. This is read-only pricing. /ticket and /triage do the acting.
• A DON'T is a real answer and often the most valuable one. Don't inflate value to avoid delivering it.
• If the honest answer is that the proposal is the wrong shape — the problem is real but the proposed fix isn't the right one — say that, price the right shape instead, and flag that you've changed the question.