embedded-captions

Add captions or subtitles to an existing single-subject talking-head video without editing the footage. Use for plain verbatim captions, cinematic captions embedded behind the subject, VFX captions, “炸/特效/酷炫字幕,” or a named identity from the 35-style catalog. Route by visual identity, not by backend engine. The quiet `anchor` rail is the default; embed every word only when the user explicitly wants a fully cinematic treatment. The workflow runs locally end to end, including transcription and subject matting; split multi-shot footage before applying it.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

77%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

An impressively actionable, well-gated pipeline with concrete commands, validation checkpoints, and feedback loops. Its weaknesses are verbosity (repeated mantras and dense prose) and broken progressive disclosure: the most important referenced routing files (CATALOG.md, themes/README.md, dna/README.md) are missing from the bundle, so the clearly-signaled navigation leads nowhere for those paths.

Suggestions

Ship the missing load-bearing references (CATALOG.md, themes/README.md, themes/*.json, dna/README.md) or inline the routing essentials they carry — the body repeatedly defers to them as the 'single source of truth', so their absence breaks navigation.

Tighten the prose: the rail/embed model and 'embedding every word is the common mistake' are restated in the TL;DR, Caption model, and Non-negotiables — state each once and link; break the long compound sentences in Non-negotiables into shorter rules.

Add one concrete plan.json / theme.json example inline so the single creative step ('author a small JSON of creative choices') is self-contained instead of pointing only to a script-header schema.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and it assumes Claude's intelligence (no explanation of what captions/ffmpeg are), but it is dense with very long compound sentences, heavy parentheticals, and repeated mantras ('embedding every word is the common mistake' appears three times; 'embed is the scarce, earned peak' is restated across the TL;DR, Caption model, and Non-negotiables) that could be tightened.	3 / 5
Actionability	Fully executable guidance throughout — exact probes ('ffprobe', 'ffmpeg ... -vframes 1'), the 5-step pipeline ('bash scripts/prepare.sh <project>', 'node scripts/preview-frames.cjs <project>', 'render-and-composite.sh'), and concrete field lists for the authored JSON ('dna', 'lines (verbatim, transcript order; 1-5 words each)', 'minors', 'hero:{match}') cover the common cases copy-paste ready.	5 / 5
Workflow Clarity	A clearly sequenced 5-step pipeline with explicit validation checkpoints — a 'Decision gate — RUN FIRST' refuse-list, Visual QA previews 'BEFORE you render', check-timing/overflow/occlusion gates, and a feedback loop ('recompile, re-preview — each loop costs seconds. Render once, when the previews pass') backed by checklists (Non-negotiables, the 5-point QA list).	5 / 5
Progressive Disclosure	Structure and signaling are good — a 'Shared knowledge' table maps each doc to its purpose and references are one level deep — but load-bearing referenced files are absent from the bundle: CATALOG.md (called 'single source of truth for routing' and 'the whole answer space', referenced ~10×), themes/README.md and themes/*.json, and dna/README.md do not exist (the themes/ and dna/ directories are missing), so navigation to the central routing material is broken.	3 / 5
	Total	16 / 20 Passed

Description

87%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, niche description that explicitly pairs a concrete 'what' with a 'Use for...' trigger clause and includes natural user phrases (including non-English '炸/特效/酷炫字幕'). It is slightly held back from a perfect mark by jargon ('35-style catalog', 'anchor rail', 'backend engine') and routing guidance that pads the capability statement.

Dimension	Reasoning	Score
Specificity	Names the domain (single-subject talking-head video) and several concrete capabilities — verbatim captions, cinematic embeds behind the subject, VFX/'炸' captions, transcription, and subject matting — but interleaves behavioral routing guidance ('Route by visual identity, not by backend engine') that reads as workflow direction rather than a clean capability list, leaving minor coverage gaps.	4 / 5
Completeness	Explicitly answers both: the 'what' ('Add captions or subtitles to an existing single-subject talking-head video without editing the footage') and the 'when' via a concrete 'Use for plain verbatim captions, cinematic captions embedded behind the subject, VFX captions...' trigger clause.	5 / 5
Trigger Term Quality	Strong natural keyword coverage with synonyms ('captions', 'subtitles') and genuine user phrases ('VFX captions', '炸/特效/酷炫字幕'), but it is diluted by internal jargon a user would not say ('named identity from the 35-style catalog', 'anchor rail', 'backend engine').	4 / 5
Distinctiveness Conflict Risk	A clear niche — captions embedded behind a talking-head subject via matte occlusion, with VFX/'炸' and named-identity triggers — that is unlikely to fire for any other skill.	5 / 5
	Total	18 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
relative_links	Relative link issues: 12 missing	Warning

	Total	15 / 16 Passed

Repository: heygen-com/hyperframes
Path: skills/embedded-captions/SKILL.md
Commit: 0f8eb89

Reviewed: 6 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.