Heads up seven up
Classify inputs as allowed, rejected, or quarantined by looking beneath filenames, labels, or surface appearance at provenance, structure, content indicators, and behavior. Use only when the user explicitly invokes Heads Up Seven Up for defensive input review.From its SKILL.md
npx -y skills add jawsublime-byte/simon-says --skill heads-up-seven-upAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 27 days oldThe repository was created 27 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.3 KB, 212 tokens by cl100k_base, as published. Nobody here has run it
Heads Up Seven Up
Do not identify safety by looking at the face value alone.
Inspect
- Define the receiving boundary and its allowed input contract.
- Record provenance, acquisition path, identity, declared type, observed type, size, encoding, structure, and relevant content indicators.
- Check mismatches, polyglots, malformed structure, hidden payloads, decompression growth, external references, active content, and parser-specific risk when relevant.
- Use isolated inspection for untrusted inputs.
- Classify:
- allow: contract and checks pass;
- reject: a defined rule fails;
- quarantine: evidence is incomplete, contradictory, or requires deeper analysis.
- Record the evidence and policy rule behind the decision.
Never execute an unknown input merely to identify it. Never treat an extension, MIME label, sender name, or model guess as proof.
Completion
Finish when every input has a decision, evidence, and next action. Uncertainty must become quarantine, not a blind safe verdict.
What ships with it: 1 file
247 B alongside SKILL.md
agents/
- openai.yaml247 B