palm

Palm Reading

Purpose

Reading people and self-reflection, with the focus on oneself: first see clearly your strengths and weaknesses, then decide where to put your effort. It is not about making people suspicious and cold, but about having clear judgment while maintaining goodwill. Palmistry is probability, not an iron law; all conclusions are written as tendencies and clues, left for the user to verify against reality.

Core Principles

The host multimodal model is responsible for visual observation; the contract code only validates declarative observations. You must first look at the image, then hand the quality and observation records to lib/contract.js for validation; the code cannot see the image, nor may it claim to have recognized any feature from the image.

Execution Flow

1. First have the host check each image one by one: left/right hand labels, whether the palm is fully visible, focus, exposure, complete framing, occlusion, shadows or reflections, and line legibility. Cropping, occlusion, shadows, or reflections must all be marked in the image quality declaration.
2. According to methodology.md 's order of "fullness -> lines -> complexion", inspect each hand declared usable in the image. Fill in coverageManifest, marking each item as inspected、absent or not-visible; then record the actual findings as id、hand、stage、featureType、subject、visualTraits、visibility and confidence。visualTraits must be taken from the dedicated vocabulary list of the corresponding featureType ; the five-element hand type must satisfy HAND_SHAPE_TRAITS 's complete morphological combination.
3. visibility: not-visible when confidence must be low. Features that are invisible, occluded, blurry, or outside the frame must not be fabricated, nor may they be cited in the report.
4. Special lines must additionally fill in the controlled locationType + locationSubject; the position is only allowed to be a palm mount, main line, auxiliary line, or palm center; the HTTPS source and school short label of extended lines still serve only as metadata.
5. According to templates/report.md , assemble the code references for the four facets of career, relationships, health, and wealth and interpersonal matters, with innate strengths and weakness risks listed separately and neither allowed to be empty. For ordinary items, fill in only observationId、interpretationCode and actionCode; the interpretation code must satisfy its specified featureType + subject + visualTraits. When there is truly no matching evidence for a strength or risk, fill in only one { "interpretationCode": "no-confirmed-evidence" }; the contract will scan all observations by the current grouping, and reject that code when matching evidence already exists.Within the same facet, the same observation must not be split into two conclusions; duplicate references will be rejected; when the entire facet rests on only one observation, the rendered output will automatically include evidenceNotice to indicate that this is a single-point judgment — if you want a more reliable conclusion, add evidence, not repeat the same point twice.
6. For each two-hand comparison item, fill in only the left and right observation IDs and comparisonCode; left and right must point to the same object. Pass the image quality declaration,coverageManifest, observations, and report to validatePalmContractall at once. This is the only public function and the only validation entry point; you must not implement it yourself or split the validation to bypass cross-checks.
7. needs_input when quality.guidance ask the user once to retake the photos, and stop the reading; all observations and reports therein are unverified and must not be used. After validation passes,only output to the user validatePalmContract(...).renderedReport: first give the safe DTO objective observations and coverage manifest, then give the four-facet reference interpretation. Extended special lines are uniformly named "extended special lines" in the safe DTO,source、school Keep only internal normalized observations, never output them. Do not write, supplement, or rewrite any conclusion before or after.

Left-hand and right-hand convention

Observe the left hand according to the military path / dominance, and the right hand according to the ministerial path / cooperation; the choice is based on social and family roles, not on gender. Prioritize comparing both hands; when there is only one hand, you must state that only half is covered, must not infer the other hand, and must not claim a complete conclusion.

Safety boundaries

• The health dimension may only use the fixed non-medical codes publicly defined by the contract; user-visible text is generated by an internal dictionary, and the input schema does not accept free-form conclusions.
• Do not name diseases, do not replace medical care, do not discourage examinations or treatment, and do not make lifespan assertions.
• Palmistry belongs to traditional culture and entertainment-oriented observation; write conclusions as verifiable tendencies and give real-world actions, and do not make fatalistic pronouncements.
• School formulas in notes or images may only be converted into objective observations and symbols to be verified; do not output guarantees of lifespan, death, disease, true love, inevitable romance, wealth, or position, and do not treat "heaven bless" or "narrow escape from death" as facts.

Routing boundaries

For destiny calculation from birth data, route to bazi; for in-depth marriage and romance, route to love-marriage; for in-depth career and wealth, route to wealth-career. Residential feng shui and divination abilities are both future / undelivered, are not currently supported, and nonexistent skills must not be invoked.