FORMWEFT CLEAN ROOM · CREATIVE EVIDENCE
Extract verifiable variables from public samples, record sources, timing, and selection bias, then run controlled experiments with your own product assets.
FIVE EVIDENCE STEPS
High sales, top ranking, or high engagement doesn't mean a specific main image caused the result; any pattern is first a hypothesis, not a fact.
QUESTION
Limit platform, category, price range, time window, and the decision to be solved, e.g., whether the first image should prioritize showing the product itself or the usage result, to avoid aimless collection.
Leave behindSample and research contract
ConfirmQuestions can be answered through follow-up experiments.
SAMPLE
Only record publicly accessible information, source links, observation time, and page position; do not download or republish protected images; do not treat rankings as sales evidence.
Leave behindTraceable source sample index
ConfirmSample acquisition and recording methods comply with rules.
CODE
Describe samples with repeatable fields such as subject area, angle, background complexity, people, text, benefit points, evidence format, and color contrast.
Leave behindVariable dictionary and coding table
ConfirmTwo reviewers have similar judgments on the same sample.
HYPOTHESIS
Distinguish observed facts, possible explanations, and unknown factors; state applicable conditions, counterexamples, and needed data; do not claim to have found a universal viral formula.
Leave behindEvidence cards and hypothesis list
ConfirmEach hypothesis can be falsified.
EXPERIMENT
Change only one main variable at a time, preset primary metrics, guardrail metrics, sample thresholds, and stopping conditions; keep failed versions and review conclusions.
Leave behindExperiment brief and results log
ConfirmResults are reproducible and do not infringe on third-party assets.
FIXED RELEASE
FORMWEFT clean-room original documentation Skill. MIT covers only the original documentation in this package; current fixed hash has passed structural validation and static review. External sample access only after user approves specific domains.
You can download this reviewed version's resource package. If it includes SKILL.md, provide it and the related documentation to your AI tool; first read and check the materials and usage boundaries, and do not run anything automatically.
You can also select the text to copy directly; read it first before executing.
Rewritten independently based on the Feishu directory topic clean-room, without copying or redistributing the original package text. Conclusions apply only to the current fixed hash: the archive structure has been parsed, static rule result is eligible, with 6 files and no scripts, dependencies, or install hooks; online boundaries follow the statements in this resource, and files are output only to user-approved locations. Static review cannot replace runtime environment, input data, or later-version rechecks.
FORMWEFTcc2c0f406957474be03c52b5e348b6ff30b5aebb84ebbc878e36b7022598a024Describe your category, market, channels, asset library, experiment traffic, and decision cadence, and we can continue to build sample libraries, experiment dashboards, asset management, and review mechanisms.
The team reviews your request and follows up by email. Scope, fees and timing are agreed separately; submitting does not create an order.
