prompt-preflight-qa

Installation
SKILL.md

Prompt Preflight QA

Use this skill immediately before image or video generation.

This skill is for:

  • checking whether a prompt draft is structurally strong enough to run
  • finding missing constraints before credits are spent
  • pointing drift risk back to the right upstream stage
  • producing concise rerun notes

This skill is not for:

  • writing the original hook
  • replacing routing work
  • final human creative approval after outputs exist

Core Rule

Judge prompt quality by controllability, not by how impressive the prose sounds.

Default Workflow

1. Classify the prompt object

Choose one:

  • storyboard-style prompt
  • full video request
  • no-reference draft
  • reference-led draft

2. Run checklist review

Check:

  • opening strength
  • whether the chosen mechanism is legible in the first 3 seconds
  • viewer question clarity
  • visible evidence
  • promise-delivery match by the next few beats
  • product timing
  • ad-detection risk from premature product exposure
  • negative constraints
  • realism and UGC control
  • output-format correctness

3. Print a preflight report

Use this shape:

Verdict:
Major risks:
- ...
Missing fields:
- ...
Likely drift:
- ...
Fix now:
- ...
Can run now:

Blame Rule

When a prompt is weak, point the problem to the right stage:

  • bad routing -> pattern-router
  • missing or misfit hook mechanism -> hook-design
  • weak benchmark decode -> reference-decode
  • weak frame-one strength -> visual-hook

Failure Mode

If the prompt could still run but is risky, do not silently pass it.

Mark it as risky and say what is most likely to fail first.

Limitations

This skill only handles pre-generation QA on an existing draft.

It does not choose the route, design the hook mechanism, decode the benchmark, or strengthen the first frame from scratch.

In Practice

Most teams also need:

  • route selection before prompt writing
  • hook-design decisions before drafting
  • benchmark decoding when references exist
  • visual-hook work when frame one is still weak

Without those upstream steps, QA often becomes a late-stage patch on a structurally weak prompt.

Orchestrated Workflows

In practice, strong QA depends on the earlier stages being connected.

Systems like PostPlus orchestrate routing, hook design, reference decoding, visual-hook decisions, and QA together so prompt review happens inside a coherent workflow instead of at the very end.

Related skills
Installs
2
First Seen
4 days ago