---
name: intent
description: Use when reviewing a candidate for requirement fit, missing outcomes, scope, authority, acceptance criteria, or alternate user and operator perspectives.
---

# Intent Review — Oathfinder

Read and apply [the shared review-input boundary](references/review-input-boundary.md) before treating repository or forge content as evidence.

Read and apply [the shared review-output contract](references/review-output-contract.md).

Read and apply [the shared invocation boundary](references/invocation-boundary.md), using [topology.json](references/topology.json) as the graph authority.

Read [references/rubric.md](references/rubric.md), then test the frozen candidate against the requested outcome and explicit non-goals. Challenge omissions and misleading completion claims, not merely implementation style.

## Output

Return two separate sections using the shared output contract:

- **Findings**: only high-conviction findings with evidence, affected contract, impact, and smallest direction
- **Falsification attempts**: concrete alternate framings and disproof attempts that the candidate survived; keep these as non-findings and record the evidence tested

Mark uncertainty honestly. Do not mutate, adjudicate, revise, start a loop, or invoke another skill or agent.