Driftproof
Menu

Read the test. Try the question.

media-use

The prompts and scoring guides used to test media-use, with each report’s findings and limits.

heygen-com/hyperframes

Agent Media OS for a HyperFrames project. Resolve BGM, SFX, image, icon, brand logo, voice, color grade, or LUT into a frozen local file or paste-ready block + ledger record (one verb, `resolve`); generate via TTS / music / image models when the catalog misses; produce voiceover, transcription, captions, and background removal through one shared audio engine; operate on media (cut / reframe / transform); and reuse assets across projects. Also use for vague feedback that real footage looks dark, flat, boring, should feel retro/camcorder/print/ASCII, needs privacy, or needs a media reveal. When the host app provides its own music or sound-effect tools, use those for music and sound effects; `resolve --type bgm|sfx` needs the heygen CLI. When `HEYGEN_API_BASE` is set, HeyGen calls go through that host with no CLI sign-in.

About this skill and its measurements

The task

reuse

Link to this task

Written responses only. No tool execution, installed dependencies, media generation or cloud services are tested.

What we asked

Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. A project needs a short chime. A local candidate has the right license and duration but its fit has not been checked. Explain the next asset workflow.
The scoring guide
Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source heygen-com/hyperframes/skills/media-use/SKILL.md at c4440dfb200def5e0c0afee938778f05b5985039; SKILL.md SHA-256 0f08ff6b493a46e0505dcc9c0c607e517e0eb70e7dc2b7b5c1f51b4f27bb6181; README SHA-256 3e8f26d5e35c6a8c32c661bf1439443d8eee332eedc9bf1a32bfe3dc417c218e. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Inspect reusable candidates before generating 2. Evaluate fit rather than reuse blindly 3. Resolve a concrete local asset and preserve its record 4. Do not claim an audio sample was heard

Pass threshold: 0.75.

Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call.

Try it yourself

Paste the prompt into your own assistant, with or without the skill. Use the scoring guide to read the answer.

One try scored by hand is not a measurement.

Report 017: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.

Report 018: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.

Copying stays in your browser. Copies are never stored, counted or sent anywhere by Driftproof.

What happened on this task

Open a report to read its comparison and limits together. Beating the fake is not beating no skill.

Claude Sonnet 5.5Report 017 · Read the task result

Against the same-length fake

reuse (clearly higher): with the skill 0.85 plus or minus 0.11547; with the same-length placebo 0.6 plus or minus 0.05.

Against nothing

reuse (clearly higher): with the skill 0.85 plus or minus 0.11547; without it 0.607143 plus or minus 0.113389.

  • What we tested. Report 015's twenty skills on Sonnet 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5.
  • Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
  • What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for Report 015's twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Sonnet 5.5 judged by Claude Haiku 5.5.

Driftproof collected several answers per arm. A judge model scored each answer 5 times.

Claude Haiku 5.5Report 018 · Read the task result

Against the same-length fake

reuse (clearly higher): with the skill 0.675 plus or minus 0.086603; with the same-length placebo 0.516667 plus or minus 0.028868.

Against nothing

reuse (too few answers to tell): with the skill 0.675 plus or minus 0.086603; without it 0.5375 plus or minus 0.085391.

  • What we tested. Twenty skills on Haiku 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5, the same model.
  • The judge is the target's model. The judge uses the same model as the target. Self-grading may favor its own model's answers.
  • Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
  • What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for the twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Haiku 5.5 judged by Claude Haiku 5.5.

Driftproof collected several answers per arm. A judge model scored each answer 5 times.

The task

grade

Link to this task

Written responses only. No tool execution, installed dependencies, media generation or cloud services are tested.

What we asked

Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. The supplied frames are uniformly too dark. The user asks for ordinary exposure correction, with no LUT supplied. Propose the review and persistence workflow.
The scoring guide
Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source heygen-com/hyperframes/skills/media-use/SKILL.md at c4440dfb200def5e0c0afee938778f05b5985039; SKILL.md SHA-256 0f08ff6b493a46e0505dcc9c0c607e517e0eb70e7dc2b7b5c1f51b4f27bb6181; README SHA-256 3e8f26d5e35c6a8c32c661bf1439443d8eee332eedc9bf1a32bfe3dc417c218e. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Inspect evidence before selecting a correction 2. Use a bounded correction candidate 3. Persist the treatment settings 4. Avoid inventing a LUT merely to encode exposure

Pass threshold: 0.75.

Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call.

Try it yourself

Paste the prompt into your own assistant, with or without the skill. Use the scoring guide to read the answer.

One try scored by hand is not a measurement.

Report 017: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.

Report 018: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.

Copying stays in your browser. Copies are never stored, counted or sent anywhere by Driftproof.

What happened on this task

Open a report to read its comparison and limits together. Beating the fake is not beating no skill.

Claude Sonnet 5.5Report 017 · Read the task result

Against the same-length fake

grade (too few answers to tell): with the skill 0.9 plus or minus 0; with the same-length placebo 0.92 plus or minus 0.083666.

Against nothing

grade (too few answers to tell): with the skill 0.9 plus or minus 0; without it 0.89 plus or minus 0.074162.

  • What we tested. Report 015's twenty skills on Sonnet 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5.
  • Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
  • What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for Report 015's twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Sonnet 5.5 judged by Claude Haiku 5.5.

Driftproof collected several answers per arm. A judge model scored each answer 5 times.

Claude Haiku 5.5Report 018 · Read the task result

Against the same-length fake

grade (clearly higher): with the skill 0.975 plus or minus 0.05; with the same-length placebo 0.83 plus or minus 0.083666.

Against nothing

grade (clearly higher): with the skill 0.975 plus or minus 0.05; without it 0.79375 plus or minus 0.111604.

  • What we tested. Twenty skills on Haiku 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5, the same model.
  • The judge is the target's model. The judge uses the same model as the target. Self-grading may favor its own model's answers.
  • Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
  • What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for the twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Haiku 5.5 judged by Claude Haiku 5.5.

Driftproof collected several answers per arm. A judge model scored each answer 5 times.

Words with a dotted line link to the glossary, where each is explained in plain words.