cat-mode: Visual Proof authenticity defaults (Expected before capture)

[EdbertChan]

Summary

Agents prove UI claims with screenshots. Reviewers trust those shots only when they show the same surface the claim names.

The failure mode is a near-neighbor capture: another product's login page, a generated text slide, or a mock screen treated as real UI.

This change adds standing authenticity rules: match proof to claim, write expected predicates before capture, reject synthesized UI, and require real inspection.

Review Claim

Visual proof for a claimed product surface must come from that surface's real pixels, after expected predicates are written, and a marker-only inspection line does not count as a check.

Review Lane

behavior

Review Unit

corpus-lesson

Safety Invariant

This slice only adds standing verification rules and colocated trigger scenarios to a personal skill. It does not change runtime hooks, install wiring, or product capture tooling. Agents that already follow visual-proof correctly keep the same path; the new text forbids substituting a different surface or a synthetic image as proof.

Slice Rationale

One review unit under the personal skill. The authenticity defaults belong with the user's verify posture, not a rewrite of the shared visual-proof skill.

Non-goals

• Do not change the shared visual-proof or prove-it skills beyond cross-links.
• Do not add hooks or PR-body validators in this slice.
• Do not implement Slack product UI or copy-button behavior.
• Do not merge this PR.

Assumptions

• Safety Invariant above is proposed under mandated non-interactive automate-me from reflect; treat as unconfirmed until a human edits or approves.
• Lane is behavior (product-unit corpus-lesson) rather than policy/docs.

Test Plan

Test Plan

• python3 scripts/ci/check_no_dated_provenance.py --base origin/main
• python3 -m unittest tests.test_cat_mode.TestVisualProofAuthenticity -v
• python3 scripts/ci/check_skill_test_coverage.py --base origin/main --head HEAD
• python3 scripts/ci/check_skills_three_harnesses.py
• python3 scripts/ci/check_ecosystem_boundaries.py
• python3 engine/skills/make-pr/scripts/preflight.py --base origin/main --body-file /tmp/catstack-automate-me-visual-proof-pr-description.md → ok preflight passed

Revert Plan

Revert Plan

• Safe to revert? Yes
• Revert command: git revert <sha>
• Post-revert steps: None (re-run install if a machine had already pulled the skill)
• Data migration? No

---

Note

Low Risk
Corpus and unittest changes only; no hooks, install wiring, or capture tooling—behavior guidance for agents already following visual-proof correctly.

Overview
Adds standing Visual Proof authenticity rules to personal cat-mode (Verify section in SKILL.md plus expanded references/verify.md), extending [[visual-proof]] without changing shared skills or hooks.

Agents must match capture surface to the Review Claim, declare Expected surface/predicates (and per-case lists for OR claims) before capture, treat marker-only Manually inspected: as not a check, reject synthesized UI (slides, HTML mocks, etc.) unless the user asked for a mockup, and cover each major behavioral case with its own proof or a named waiver.

Four colocated skill test scenarios document fire vs. stay-silent cases; TestVisualProofAuthenticity in tests/test_cat_mode.py locks key phrases in the skill and reference.

^{Reviewed by Cursor Bugbot for commit 86a58ad. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_0b70c110-1980-4e27-ae45-e93a0a2b2b93)

[mergify]

Queued — the merge queue status continues in this comment ↓.

[EdbertChan]

@Mergifyio queue

[mergify]

Merge Queue Status

• ✅ Entered queue — 2026-09-29 00:39 UTC · Rule: admin-bypass · triggered by @EdbertChan with the @mergifyio queue command
• ✅ Checks passed · on draft merge queue: checking #1207 on main (0aa9224) #1217
• ✅ Merged — 2026-09-29 01:24 UTC · at 3902aa5c1f5c639ed6480f4cb5dfb37eb673bd4b · squash

This pull request spent 45 minutes 6 seconds in the queue, including 44 minutes 39 seconds running CI.

Required conditions to merge

• check-success = lint
• check-success = test
• check-success = validate