ADFOUNDRY / ACADEMYCourse map

A7 · 50 minutes

Review, diagnose, and repair

Outcome: Produce an evidence-based verdict across technical, craft, brand, and honesty dimensions, then perform one bounded repair.

Objectives

  • Separate observed defects from preferences and record every note at a timecode.
  • Score technical, craft, brand, and honesty dimensions without averaging away a critical failure.
  • Repair the failed dimension with one controlled change while keeping accepted work locked.

Learn one thing

Observation before diagnosis

A useful note states what is seen or heard and where it occurs. The proposed cause comes second, because the same symptom can come from generation, edit, or mix choices.

Choose: Write timecoded evidence before suggesting a fix.
Critical gates do not average

A beautiful ad with a false claim is not mostly approved. Identity drift, unreadable output, missing rights, or unsupported proof can fail the artifact regardless of its total score.

Choose: Mark honesty and rights failures as stop conditions.
Repair one dimension

A bounded retry names the failed dimension, changes one cause, and locks accepted traits. Rebuilding everything destroys evidence about what improved.

Choose: Choose one repair and one retry limit before reopening a tool.

Make the artifact

  1. Run cold passes

    Watch full-speed, muted, audio-only, and frame-by-frame passes without changing the edit between passes.

    Check: Each pass produces timecoded observations rather than general reactions.
  2. Score four dimensions

    Complete technical, craft, brand, and honesty scores and explicitly mark any critical stop condition.

    Check: A strong total cannot conceal a zero in honesty or identity.
  3. Name one cause

    Select the failed dimension, cite its evidence, propose one causal change, and list every accepted trait to lock.

    Check: The repair request changes one bounded variable.
  4. Repair and compare

    Make at most one planned retry, compare old and new versions side by side, and record approve, repair, or reject.

    Check: The new verdict states whether the targeted evidence improved.

Copy, then customize

Prompt skeleton

Audit these TIMECODED OBSERVATIONS against the VERIFIED BRIEF, REFERENCE LOCKS, and RUBRIC. Return four separate sections: TECHNICAL, CRAFT, BRAND, HONESTY. For each, cite only supplied evidence, assign 0–2, and distinguish observation from inferred cause. Flag unsupported claims, missing rights, identity drift, or deceptive realism as stop conditions. If repair is allowed, propose exactly one changed variable, a list of traits to keep locked, and a one-retry acceptance check. Do not average away a zero in honesty. INPUTS: [paste].

Never upload confidential media for automated review without authority, and never let a high craft score override consent, rights, or truth failures.

Tools, clearly labeled

Local plain-text fileslocal · beginnerCompleting the supplied Markdown and JSON worksheets without a cloud accountAffiliate: no · Official site ↗Frame.iosubscription · beginnerCollecting timecoded review notes and marking a bounded repair completeAffiliate: no · Official site ↗

Downloads

Check / lock

Pass when the verdict is evidence-based and the approved version clears every critical honesty, rights, identity, and technical gate.

Rubric — 0 / 8