Brand Safety Audit
Live · Buffer data

Brand Safety Audit

SWEAT440 network · Instagram brand safety audit · Caption + visual creative analysis — human review required on all flagged items

How to use this tool
  1. Install the Claude skill (one time only). Download it here, then in Claude.ai go to Settings → Capabilities, turn on Code execution and file creation, and under Skills click Upload skill and pick the downloaded .zip. The skill teaches Claude the 6 evaluation categories and the exact file format this dashboard needs. You also need the Claude in Chrome extension installed and connected, since that's how Claude looks at each post's image.
  2. Download the posts. In the box below, the from date is already set to the day after the last reviewed date. Pick the to date and click Download posts. You get a .json file with every feed post (Static and Carousel) and every Story from all studios in that range. Reels aren't included yet.
  3. Run the audit in Claude. Open a new chat in Claude.ai with Claude in Chrome connected, attach the posts file, then copy and paste this prompt:
    Run the brand safety audit on this file using the skill sweat440-brand-safety-audit
    Claude reads each post's caption, opens its image in Chrome to look at it, and gives you back an audit .json file to download. Stories are reviewed from the copy Buffer keeps, which stays available after the Story expires on Instagram; for video Stories only the cover frame is kept. Large ranges take a while. If the chat stops midway, ask Claude to continue, and it picks up where it left off.
  4. Publish the result. In Publish Claude's result, choose the audit file Claude gave you and click Publish. It's saved for everyone and the Data reviewed date moves forward.
  5. Review the flags. Pick a date range under Filter by post date and click Apply. Each flag says where the problem was found (caption or image) and why it was flagged, with a suggested action. Open each flagged post on Instagram and click its status to move it from Pending review to Reported to Ops or Cleared. Then click Save status changes so everyone sees it. The period summary under the date filter counts accounts, clean accounts and flags for that period.
Flags are a triage queue, not a verdict. A person decides on every flagged post. See the Evaluation Criteria tab for what each category covers.
Data reviewed: —
Step 2 · Get posts from: to
Step 4 · Publish Claude's result:
Filter by post date (all cycles): to
Pick a date range above and click Apply to see posts from that period.
This audit evaluates SWEAT440 Instagram content against 6 brand-safety categories. It is a triage tool — its output is a flag-and-review queue for human adjudication. It surfaces genuine risk, not aesthetic preferences. The only strict visual rule is logo integrity. Use the status button on each flag to track its lifecycle: Pending review → Reported to Ops → Cleared. Status changes are kept in this session until you click Save status changes at the top — that's what actually publishes them for everyone.
HIGH
Clear, unambiguous violation. Logo recolored or distorted; explicit lyric soundtrack confirmed; active alcohol consumption on camera; overt political endorsement. Immediate action.
MEDIUM
Likely violation, context-dependent. Suggestive pose; unconfirmed explicit audio; external partnership without documented approval. Review this week.
LOW
Borderline — quick human glance. Cultural/social messaging short of a violation; loose brand formatting. Confirm or dismiss.
The 6 Evaluation Categories
CategoryFlag if you see thisDo NOT flag thisTypical confidence
1. Political or religious messagingCandidate endorsement, partisan statements, proselytizing content, divisive legislation stancesCultural holiday acknowledgment, general inclusivity language, community appreciationHIGH if overt · MED if ambiguous
2. Explicit lyrics in audioReels where the soundtrack contains confirmed explicit lyricsEnergetic music without confirmed explicit content; instrumentalsHIGH confirmed · MED suspected
3. Overly sexualized contentDeliberate sexualization beyond fitness norms; suggestive posing unrelated to exerciseAthletic bodies, gym wear, sports bras, sweat, exertion — standard fitness contentMED or LOW — gradient
4. Active alcohol consumptionAlcohol actively consumed on camera; "happy hour" language with alcohol emojisBackground presence only; non-alcoholic beverage partnershipsHIGH if shown · LOW if presence
5. Logo misuseAny modification of the official mark: recoloring, distortion, stretching, unofficial fonts, unapproved backgroundsOfficial logo used correctly with approved colors and backgroundsHIGH — binary, zero tolerance
6. Loose brand adherenceUnapproved external partnerships in-studio; third-party medical providers without documented approval; major brand voice departureMinor caption formatting; emoji use; slight tone variation; aligned community eventsMED or LOW — light touch
SWEAT440 · Brand Safety Audit · Confidential — Internal Use Only Triage tool only — human review required on all flagged items