challenger

Devil's advocate against a freshly drafted implementation plan — the paper stage, before you approve it. Attacks the approach with codebase evidence: concrete failure scenarios and cheaper alternatives, never vague doubt.

Triggered by	/ticket:pick step 3, right after the plan is drafted, before the Plan gate — dispatched alongside the plan so you judge both together
Also usable	Standalone, against any design or plan
Tools	Read, Grep, Glob, Bash

Flow

flowchart LR
    In1["Ticket body +<br/>'## Decisions & assumptions'"] --> A["challenger"]
    In2["Drafted plan"] --> A
    In3["Diff base"] --> A
    A --> Explore["Explores code:<br/>Read / Grep / Glob,<br/>git log via Bash"]
    Explore --> Out["Steelman + verdict +<br/>up to 3 evidence-grounded challenges"]
    Out --> Gate["Plan gate<br/>(you decide)"]

Verdicts

Verdict	Meaning
PLAN STANDS	No material weakness found
WEAKNESSES	Named, evidence-grounded concerns — plan may still be approved as-is
RIVAL ROUTE	A cheaper or more direct route exists
ASSUMPTION-BROKEN	A load-bearing assumption in the plan doesn't hold

Challenges are capped at 3, each grounded in one of: hidden coupling, a concrete failure scenario, a cheaper route, an irreversible step, a load-bearing assumption, or an effort mismatch.

Gates directly? No — feeds your Approve / Edit / Abandon decision at the Plan gate; it never edits or blocks on its own.

See also

• code-challenger — the loop-time sibling that attacks the code as built, instead of the paper plan