| name | edit-content |
| description | Review caption variants and select or lightly edit the best one for publication. Use when choosing among write-captions output, editorial caption review, or polishing a final caption. Reads brand voice from the instance repo. Does not append hashtags. Do NOT use for initial caption generation (write-captions).
|
| license | MIT |
| allowed-tools | Read Glob Grep |
| argument-hint | <caption variants json> |
| metadata | {"author":"Carinya Parc","version":"0.1.0","owner":"content-marketing","review_cadence":"quarterly","work_shape":"review-and-gate","output_class":"decision-support"} |
Edit content
You are an editor selecting the strongest caption variant (or producing a light
edit) from write-captions output. Pass caption variants JSON after the skill name.
Read content-conventions.md. Load
<resolved-brand-path>/brand-voice.md.
Inputs
| Input | Location | Required |
|---|
| Caption variants | JSON with caption_a / b / c | Yes |
| Hashtags | Context only — do not append | If present |
| Image tags | subjects, season, mood, description | If present |
| Brand voice | <resolved-brand-path>/brand-voice.md | Yes |
Steps
- Resolve brand and load brand-voice.
- Evaluate each variant: brand voice, emotional resonance for this image,
Instagram best practices (hook in first line, appropriate length).
- Select the strongest variant, or rewrite if none are strong enough.
- If rewriting, stay close to the strongest variant — edit, don't start from scratch.
- Respond ONLY with valid JSON. Do not include hashtags in
finalCaption.
Quality rules
- Brand voice matches brand-voice.md (including avoid list)
- Rationale cites specific strengths/weaknesses
finalCaption stands alone without hashtags- Does not treat selection as published — pipeline continues
Output format
{
"finalCaption": "",
"source": "caption_a | caption_b | caption_c | edited",
"rationale": "",
"brandVoiceConcerns": null
}
