Read the discovery protocol before acting. Bundled templates are optional aids, not a mandatory artifact checklist.
Challenge, do not take overRead the actual frame, evidence, alternatives and human decision. Use a separate review context when available. Otherwise disclose a same-context self-check; do not claim independence by adopting a role. Never edit records under review.
Check unsupported critical claims, ignored counterevidence, excluded alternatives, constraints, and appetite. Look for concrete rabbit holes that threaten the outcome. For a premortem imagine failure at a relevant future point and trace a few plausible causes grounded in current context, not exotic catastrophes.
Each material finding identifies evidence, consequence, and the smallest useful response. Separate observed defects, conditional risks, unknowns and speculation. An unsupported hypothetical or optional polish is not a blocker. Unknown experiment results do not block a properly bounded learning decision.
Save challenge.md only when useful. Record review context and CLEAR/FINDINGS/ INSUFFICIENT EVIDENCE, applicable input identities, and next check or decision amendment. Findings are advisory; the human still chooses. A clean, short review is valid. Do not approve the promise or start another stage automatically.
Compound durable learning
If this work surfaces a durable, non-obvious lesson likely to help future ITP work,
invoke the compound-learning skill before finishing. Do not invoke it for routine
facts, one-off details, already-recorded evidence, or merely because the work was
difficult. Compounding is advisory: it must not create or replace a human decision,
change an approved promise, add scope, or silently rewrite shared ITP behavior.