Skip to content

aaron-he-zhu/aaron-marketing-skills / influencer/activate/content-amplifier

content-amplifier

Verified

Use when the user asks to "amplify influencer content with paid media", "set up whitelisting or Spark Ads", "decide which posts to boost", "repurpose influencer content", "turn one video into multiple ads", or "build a UGC asset library"; produces (paid mode) a content-selection scorecard, a paid amplification strategy (whitelisting/boosting/dark posts), audience targeting, and a budget+optimization plan, or (repurpose mode) a rights-tracked content inventory, a 1-video-to-10+-asset repurposing map, per-format transformation specs, and a 30-day distribution plan. Not for gating whether a deliverable is publishable or FTC-compliant — use creator-content-auditor; not for the always-on brand posting calendar — use social-calendar-builder; not for drafting a net-new idea into platform-native packages — use social-creative-builder. 复用达人内容 / 内容放量.

FileKindSize
SKILL.mdskill md25.8 KB
references/atom-extraction.mdmarkdown5.6 KB
references/templates.mdmarkdown34.4 KB

references/atom-extraction.md

View on GitHub
# Content-Atom Extraction Method

A method for breaking one piece of UGC into reusable "content atoms" — the smallest standalone units worth repurposing. The agent does this by reading the pasted transcript, caption, or review text. No audio/video processing, no libraries: you read the words and extract.

> Method only. Do NOT install or call whisper, mediapipe, pandas, or any package. If the user has a video, ask them to paste the transcript or captions and work from that text.

**Source identity boundary**: before extraction, bind the transient source text to exactly three persistent fields: stable opaque `creator_ref`, exact frozen `approved_asset_ref`, and an opaque authorized `source_ref`. Never persist a raw handle, creator name, profile/content URL, provider ID, or a raw URL disguised as `source_ref`. If any of the three refs is missing, return `NEEDS_INPUT` and do not save or hand off atom records.

## 1. The 7 Atom Tiers

Read the source text and pull every standalone unit that fits one of these tiers. Tag each atom with a timestamp (or text position if no timecodes) and candidate formats. The tier does not itself rank, score, approve, or choose a paid/hero placement.

| Tier | What it is | Looks like | Suggested platforms |
|------|-----------|-----------|--------------------|
| `narrative_arc` | The whole before→after journey in one line | "I had X problem, tried this, now Y" | YouTube, landing page hero, case study |
| `quote` | A short, quotable line in the creator's voice | "This is the only one that actually worked." | quote card, website testimonial, ad headline, email |
| `controversial_take` | A claim that splits opinion or pushes back on common advice | "Everyone says X — they're wrong." | X, Reddit, TikTok hook, ad hook |
| `data_point` | A specific number, result, or measurable claim | "Saved 4 hours a week." "Down 12 lbs in 6 weeks." | ad copy, landing-page stat, email subject |
| `story` | A self-contained anecdote with a beginning and payoff | "So last Tuesday I…" | Reels, TikTok, Stories, blog snippet |
| `framework` | A named or numbered method the creator teaches | "My 3-step morning routine" | carousel, LinkedIn, YouTube Short, blog |
| `prediction` | A forward-looking claim about a trend or outcome | "By next year everyone will…" | X, LinkedIn, thought-leadership post |

### Per-atom record

```markdown
- atom_id: A-001
  tier: quote
  text: "This is the only one that actually worked."
  timestamp: 00:00:18   # or char-offset / "para 2" if no timecodes
  creator_ref: creator-opaque-ref
  approved_asset_ref: frozen-approved-asset-ref
  source_ref: opaque-authorized-source-ref
  candidate_formats: [quote card, website testimonial, ad headline]
  selection_rule_ref: NEEDS_INPUT
  selection_score: null
  score_state: NOT_SCORED
```

## 2. Selection Rule — No Built-In Score

This reference supplies no default “virality” weights, bonuses, cutoff, ranking, or placement priority. Extracting an atom is not evidence that it should receive paid spend or a hero placement.

Before assigning any numeric score, rank, keep/drop decision, or placement, require one of:

- a user-approved rule that explicitly defines criteria, weights, scale, cutoff, intended use, and approval ref; or
- a source-dated rule whose evidence ref, population/context, date, and intended use are compatible with this campaign.

If neither exists, set `selection_rule_ref: NEEDS_INPUT`, `selection_score: null`, and `score_state: NOT_SCORED/NEEDS_INPUT`. Preserve the extracted atoms without ranking them and do not automatically select anything for paid, hero, or another destination.

If a qualifying rule exists, copy its criteria and weights exactly, show the calculation inputs and result, and record `selection_rule_ref`. Do not add an atom-type bonus, content bonus, or unstated tie-breaker.

## 3. Near-Duplicate Evidence (No Default Threshold)

Before you publish a batch, you may calculate lexical similarity as descriptive evidence. Similarity alone does not authorize dropping or suppressing an atom.

**Jaccard similarity** = (words shared by both) / (all distinct words across both). Compute it by hand on lowercased word sets, dropping punctuation and common stop-words (the, a, is, and, to, of, it, this).

```
J(A, B) = |words(A) ∩ words(B)| / |words(A) ∪ words(B)|
```

Require a user-approved or source-dated compatible `duplicate_rule_ref` before applying a threshold or keep/drop decision. Without it, report the calculated similarity and `duplicate_state: NOT_DECIDED/NEEDS_INPUT`; do not assume a cutoff.

Check in two places:

1. **Within the current batch** — compare atoms only when requested. Apply a flag/keep/drop action only under the cited `duplicate_rule_ref`; a selection score is unavailable unless its separate rule passed.
2. **Against recent memory** — read authorized atom records from the user-supplied/source-dated lookback window. Apply an already-used flag or skip action only under the cited `duplicate_rule_ref`.

```markdown
- atom_id: A-007
  creator_ref: creator-opaque-ref
  approved_asset_ref: frozen-approved-asset-ref
  source_ref: opaque-authorized-source-ref
  text: "It's the only one that actually worked for me."
  jaccard_similarity: 0.71
  duplicate_rule_ref: NEEDS_INPUT
  duplicate_state: NOT_DECIDED/NEEDS_INPUT
  decision: none
```

Worked calculation: `"this is the only one that actually worked"` vs `"the only one that actually worked for me"` — after the declared tokenization/stop-word treatment, shared = 5 and union = 7, so **J = 5/7 ≈ 0.71**. With no approved `duplicate_rule_ref`, report only that calculated value; do not flag or drop either atom.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 196kUpdated Claude CodeCodexVerifiedInstall