# Creator Portfolio Annotation

> Annotate every observed creator post at portfolio depth using titles, covers, post copy, public metadata, and available evidence. Use to build open-ended topic, format, promise, audience-value, proof, and visual-language fields without pretending surface evidence is a full video reconstruction.

- Skill: `cyl19970726/creator-portfolio-annotation` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add cyl19970726/creator-portfolio-annotation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/cyl19970726/creator-portfolio-annotation/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: cyl19970726 (https://skillmd.com/u/cyl19970726)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/cyl19970726/creator-portfolio-annotation

---


# Creator Portfolio Annotation

Turn a normalized corpus into comparable post-level observations. Preserve the difference between surface annotation and deep video evidence.

## Inputs

- normalized creator corpus and collection status;
- local covers, post copy, detail snapshots, and any validated reconstruction references;
- annotation revision and previous labels when updating.

## Annotate every observed post

- topic and user problem;
- content form and intent;
- hook and promised payoff;
- probable audience and use context;
- value offered;
- proof mode;
- presenter, composition, aspect ratio, screen/UI role, text density, and result visibility;
- content-architecture observations supported by visible evidence;
- confidence, evidence references, conflicts, and unknowns.

Use multi-label, open-ended concepts. Create a new label when the corpus demands it; do not force a post into a fixed taxonomy. Record `unclassified` when the evidence is insufficient.

## Evidence discipline

- Title, cover, copy, metrics, comments, transcript, frames, and reconstruction are distinct evidence scopes.
- A title can identify a candidate topic or promise but cannot establish the video's actual steps, argument, directing logic, visual sequence, or causal mechanism.
- A cover can support composition and promise observations but not unshown results.
- Reuse validated deep findings through references; do not copy them into unsupported surface rows.

## Output

Produce `portfolio-annotations.json` keyed by stable post ID, with annotation revision, source revision, evidence scope, confidence, conflicts, unknowns, and per-field evidence references. Every observed corpus ID must have exactly one row, including explicitly unclassified rows.

The artifact is a pinned input to creator synthesis. It describes the full visible denominator; it does not replace
the canonical 21 comparison set or the representative deep-video set. Deterministic title/media heuristics may create
the first conservative pass, but must keep proof, visual sequence and body structure unknown until stronger evidence exists.

Return `ANNOTATION_READY` only when ID parity, evidence references, and unknown handling validate.