SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skills100x-exaggerate

100x-exaggerate

Marketing & growth

View repository

Updated 10-15 23:45

About this skill

100x-exaggerate

One-line positioning

Input a piece of script/copy plain text, output "how to exaggerate" (exaggeration technique + intensity) + "how to create contrast" (contrast type + anchors at both ends), each item pointing back verbatim to the original script sentence, with intensity calibrated by market + ceiling. Belongs to the 100x system L2 creative generation layer, corresponding to the "3.3 Exaggeration/Contrast" step.

When to trigger

The user says:

  • "Help add some exaggeration to this script" / "How do I create contrast here" / "This part is a bit flat, how can it be more attention-grabbing" / "Add a before-and-after comparison"
  • "This script's hook isn't impactful enough" / "How can I make this line sound more exaggerated but not too fake"
  • "make this script more dramatic" / "add a before/after contrast" / "how do I exaggerate this without it looking fake" / "what's the contrast angle here"
  • Or directly provide a piece of script copy and request "design exaggeration and contrast beats" / "design exaggeration and contrast beats for this"

Input

Minimum input (Category A, mandatory):source_script——Complete script/copy plain text. This skill does not accept video files, only text (same constraint as 100x-persona ).

Soft supplement (Category B, if missing use three-level degradation, see workflow.md Phase 1）：hat_level （blackhat/grayhat/whitehat）+ market(free text, such as "US"/"US"/"Spanish-speaking region"/ "general"). When both are missing,hat_level inline inference is grayhat，market handled by conservative default (emotion_reaction_hyperbole techniques are treated as US market), do not ask the user,meta.warnings record the inference process truthfully——see axioms.md Axiom 3,workflow.md Phase 1。

Optional upstream output (Category C): This skill does not declare any fields dependent on upstream skill output——if the user has already run 100x-persona/100x-search-query and attached the output, it can be referenced in passing in rationale , but schema.json completely contains no cross-skill reference fields like persona_ref/scene_ref , more 100x-persona than the optional segment_ref of

Output

For structure see schema.json：ExaggerationContrastBundle = source_script(original text echo) + meta（hat_level/market/calibration notes/warnings）+ exaggeration_beats[](exaggeration points, each containing technique closed-set enumeration + label_cn + intensity + verbatim anchor) + contrast_beats[] (contrast points, each containing contrast_type closed-set enumeration + label_cn + verbatim anchors at both ends). Optionally render one human-readable Markdown summary (exaggeration point list + contrast point list).

Core constraints (4 axioms, see axioms.md）

  1. Exaggeration techniques (5 types) and contrast types (4 types) must be selected from closed-set enumerations, no inventing new terms——the enumerations themselves are taken from the creative bridge dictionary's L1 bridge/scene types + reference corpus signal frequency, not decided on a whim
  2. The anchor of each exaggeration point/contrast point must be a verbatim substring of the original script sentence, no fabrication——same mechanism as 100x-persona evidence citation axiom
  3. Exaggeration intensity is calibrated by market + ceiling, not the more exaggerated the better——directly absorbs the US market exaggeration intensity calibration reminder built into dictionary-06 (specific wording not quoted verbatim, see axioms.md Axiom 3), butthis restriction only applies to emotion_reaction_hyperbole(emotional reaction exaggeration), this one technique, and does not broadly restrict the other 4 techniques (known limitation: the three-tier numericalization of the ceiling table itself is an original judgment for this task, verified only on health supplement category corpora in the English and Spanish markets;meta.market The current determination for a hit on the US region does segmented exact matching on the whole segment, without doing real Chinese/English word segmentation, so it cannot recognize aliases such as "US market" embedded in a longer compound word with no separator before or after—v1.1 already fixed a more serious reverse problem: whole-string substring matching before segmentation used to, for Russia/Australia/ Belarus/market strings unrelated to the US region such as "South American country" produce false positives; v1.2 then fixed a problem in the other direction: v1.1's segmentation symbols did not include the hyphen-/&, causing writings such as "US region-general"/"US & Canada" that use a hyphen or&to combine multiple market values to be missed and judged as not hitting the US region, and the market ceiling was silently opened; now-/& has also been included in the segmentation symbols—but the category of "compound words with no separator at all" is still unresolved, see axioms.md axiom 3, TODO)
  4. The two ends of the contrast must have a real gap; the same sentence is not allowed to serve as both ends—this directly corresponds to the mechanization of the core requirement of dictionary-06 Type D visuals that "the two ends must have a perceptible gap" (specific wording is not quoted verbatim, see axioms.md axiom 4) (known limitation: it can only block "literally identical", and cannot block the more hidden degradation of "semantically repeated but literally different", the same category of ceiling as 100x-persona axiom 3 TODO, see axioms.md axiom 4)

Three-stage process

See workflow.md: Phase 1 receive + validate (category A/B,hat_level/market three-level degradation determines the ceiling) → Phase 2 (candidate sentence annotation → technique/contrast type assignment → intensity grading → item-by-item self-check → batch self-check) → Phase 3 output (assemble JSON, axioms 1-4 all pass) + user-triggered rework (L1 single item / L2 calibration adjustment, calibration adjustment can only relax the ceiling by explicitly changing hat_level/market , and cannot ignore the ceiling and raise intensity merely because of the sentence "make it a bit more exaggerated"）。

Independent invocation guarantee

Category A (source_script) missing → fixed refusal script to ask again, no guessing on the user's behalf, no accepting video files as a substitute. Category B (hat_level/market) missing goes through three-level degradation (both given → use directly; only one given → infer the other inline; both missing → infer grayhat + conservative market default),no asking again, all recorded in meta.warnings. Category C (upstream skill output) has no hard dependency fields,never require the user to run other skills first。

Banned words

let me / hope / perhaps / probably / maybe / might / let's; AI customer-service tone (as an AI / I'd be happy to / feel free to); do not fabricate effects/numbers/authority endorsements/certifications/testimonials not in the script original text; do not hard-code any customer category dictionary; do not pile up exaggeration_beats/contrast_beatsjust to make up the count; do not, in order to appear "more professional", give authority_absolutism_hyperbole techniques fabricated institution/certification names that are not in the script (this skill is only responsible for annotatingwhich technique an existing claim in the script belongs to, and is not responsible forfabricating on behalf of the scriptnew claims—compliance judgment is likewise not this skill's responsibility, following the same boundary repeatedly emphasized in dictionary-06: the skit only makes objective factual descriptions, and compliance review is left to the director for manual gatekeeping before publishing (specific wording is not quoted verbatim), see workflow.md Phase 1）。

Routing (recommended, not a hard dependency)

  • Already has exaggerated/contrast points, wants to find corresponding visual reference images → search keyword type skill (100x-search-query, if already built)
  • Want to turn exaggeration/contrast points into concrete character + scene shot design → character-scene-type skill (100x-persona, if already built)
  • Want to turn exaggeration/contrast points into AI video generation prompts → prompt assembly-type skill (if already built)
  • Want to go back to the entry point and re-triage → 100x system main entry (if already built)

All of the above are "recommended on demand after output," not prerequisites; when the corresponding skill has not yet been built, this skill's independent operation is unaffected.

Factory self-check

Run node scripts/validate.js --selftest to verify.schema.json The structural-layer constraints declared by (required/additionalProperties/enum/pattern/minLength/minItems , etc.) are actually executed by scripts/validate.js within ajv(a real JSON Schema draft-07 validator,package.json declared devDependency, first npm install then run, for ajv 8.x wiring copy 100x-search-query, not the ajv6 syntax 100x-persona used in its early days), not a hand-written field enumeration. Axiom 2 (verbatim substring anchoring), Axiom 3 (market cap ceiling table lookup), Axiom 4 (contrast non-degeneracy) are three types of cross-field/cross-entry aggregation rules that cannot be expressed structurally in vanilla JSON Schema, and are supplemented by hand-written code in scripts/validate.js :

npm install                                                   # 首次使用先装 ajv
node scripts/validate.js <bundle.json> [bundle2.json ...]   # 校验产出
node scripts/validate.js --selftest                          # 跑内置回归用例

Source

This skill is constructed based on the methodology of "Creative Visual Scene Dictionary" (v3.3) + "Script Structure Formula Library · A-G Seven Types and Effect Formulas" + "Practical Case Detailed Explanations · Four Benchmark Scripts" + reference corpus frequency statistics. The three-stage process skeleton refers to the standard generation and verification structure, andscripts/validate.js 's ajv verification executes the structural-layer constraints. For the source of each criterion, the source of field design, and which are original judgments, see sources.md。