| name | challenge |
| description | Veracity challenger — adversarially challenges factual claims in high-stakes artifacts (security findings, research reports, design docs, gate decisions) with evidence-only verdicts: CONFIRMED / CONTRADICTED / UNVERIFIABLE. Every challenge cites file:line, URL, or validator output — no speculation. Auto-runs at HIGH/CRITICAL findings and Gate A/B; invoke manually to adversarially verify any output. |
Challenger
Load and follow the instructions in the challenger agent.
Usage:
/challenge — Adversarially verify the claims in a target artifact; each claim gets CONFIRMED / CONTRADICTED / UNVERIFIABLE with cited evidence/challenge --quick — Produce the challenge list only, skipping the full rebuttal cycle (low-stakes passes)
Protocol: Full rules in agents/shared/CHALLENGER_PROTOCOL.md. sdlc-lead calls the challenger automatically between Phase 2→3 and Phase 3→4; this skill is the manual entry point.
Workflow: Identify the checkable claims in the target artifact → gather evidence per claim (file:line, URL, validator/command output) → issue an evidence-only verdict per claim → list contradicted/unverifiable claims for remediation (no speculation, no new claims)
