Skip to content

agent-anti-rationalization

Agent Anti-Rationalization — Evidence-First Discipline

Section titled “Agent Anti-Rationalization — Evidence-First Discipline”

Default to PESSIMISM; require evidence.

Forbidden thoughts:

  • “Should work” / “looks fine” → produce evidence, or say “insufficient evidence”.
  • “This is unrelated” → 1-line proof, or treat as related; park it (modules/t1k-orchestration/rules/off-scope-finding-offload.md), never dismiss it or handle it inline.
  • “One more attempt” (3+ failed) → stop; question the architecture with the user.

Five more: docs/agent-anti-rationalization.md § “Pre-trim rule text”.

Verifier sub-agents (test-runners, reviewers, doctors): prepend the anti-avoidance preamble (skills/t1k-architecture/references/fork-hygiene.md §5). Agent bodies cite this file, never inline the list.

No review (reviewer agents, /t1k:review, review stages) unless the user asked for one in this conversation; tests, gates and the merger’s own diff read stay. docs/no-unrequested-review.md