SuperAPI AI Directory

crossframe

General & other

View repository

Updated 10-16 08:50

About this skill

CrossFrame

When to Use This Skill

  • Use only when the user explicitly names CrossFrame, crossframe, /crossframe, $crossframe, or cross-scale structural diagnosis.
  • Use for Chinese-canonical structural diagnosis where facts, scale, evidence, responsibility, mechanisms, and action limits must be separated.
  • Do not use passively for ordinary analysis, writing, relationship advice, public commentary, philosophy, or long-term forecasting.

Packaged Source Note

This AAS-ready copy preserves the original CrossFrame skill body below. Chinese remains the canonical semantic layer; English metadata is only for discovery, installation, and repository review.

Limitations

  • The skill body is intentionally Chinese-canonical; English metadata is for discovery and does not replace the original Chinese terms.
  • Use only after explicit CrossFrame invocation or crossframe-suite routing; do not apply it as a generic default reasoning layer.
  • It structures analysis, drafting, and review, but does not replace source verification, domain expertise, or legal, medical, or financial judgment.

If the user's task requires multiple CrossFrame parallel skills to collaborate in sequence, first read ../crossframe-suite/SKILL.md for overall orchestration; this skill is responsible for the structural diagnosis, fact boundaries, scale windows, mechanism candidates, seven-gate review, and judgment tiers within it.

Language Principles

The authoritative semantics of this skill are in Chinese.CrossFrame is merely the English dissemination name and skill id, and does not carry the authority of conceptual interpretation.

When Chinese and English may conflict, the Chinese terms shall prevail: inheritance, reflow, open assertion, scale shift, chain of responsibility, observational reflexivity, low-condition exploratory action, exit shift, do not waste love, the eight-piece set of strong judgment, local state coordinates, processual artifact boundary.

English may be used for file names, aliases, external introductions, or necessary bilingual annotations; do not forcibly translate Chinese concepts into English and then understand them back in reverse.

Core Positioning

CrossFrame is not a prompt package that "stuffs v5.0 text into the context," but an executable structural reasoning protocol.

Every use must first form an internal reasoning artifact, then output a conclusion. The conclusion may be very short, but it must not skip fact extraction, seven-gate review, mechanism candidates, judgment tiers, source-structure continuity, and the expression gate.

Mandatory Sequence

  1. Determine the user's request type: rapid diagnosis, full diagnosis, inference, open assertion, proposition verification, strong judgment, high-reflexivity object, lightweight entry for intimate relationships, healing and transfer, public institution special topic, low-condition action, high-responsibility anti-capture review, framework boundaries, lifecycle/state coordinates, progressive closed loop, potential field/autonomous dissociation, governance continuity, framework governance and falsification, AI processual artifact boundaries, weak signal/opacity check, intermediate path without institutional infrastructure, protection of subjects unable to exit, metaphor/source transparency, instrumental accessibility, observational convergence, ultra-large-scale stress test, expression translation, theoretical backend, or concept explanation.
  2. Read references/runtime-read-policy.md、references/read-routing-map.md and, when necessary, references/v5-material-selection-map.md, to determine the v5 source modules, co-reading packages, protocols, worksheets, concept cards, and templates to be loaded this time.
  3. First locate the v5 source modules; do not open large files in full. By default, only record the needed source module, keywords, V5-H, or source range; only when source anchors are insufficient, the user requests a source audit, or a high-responsibility judgment requires verification, read the relevant local portions of references/v5-source-spine.md、references/v5-section-digest-index.md、references/v5-coverage-map.md or references/v5-term-fidelity.md in a targeted manner.
  4. Read references/continuity-closure-map.md to expand the entry package's "mandatory co-reading closure"; when package descriptions, source anchors, or downgrade details are needed, then read references/continuity-bundles.md and the corresponding references/continuity-bundles/v5/<bundle-id>.md. By default, read at most 3 entry core packages + 2 adjacent auxiliary packages; this upper limit does not restrict the mandatory co-reading closure. High-responsibility, public institution, organizational handling, and public judgment must prioritize reading the seven gates, the strong judgment eight-piece set, low-power protection, evidence downgrade, and action ceiling packages and their closures.
  5. According to templates/read-state-capsule.md , generate v5-read-state-capsule: first list source modules, then list entry packages, mandatory co-reading closure, adjacent candidates, source anchors, downgrade boundaries, and downstream reading strategy. The suite does not generate capsules; capsules are generated by this core layer and passed to specialized skills, essay, and review.
  6. Fill in the internal intake: object, scale, facts, evidence gaps, user purpose, affected objects, observational impact, power structure, action ceiling.
  7. Pass the seven gates: object gate, evidence gate, scale gate, responsibility gate, observation gate, power gate, action gate. If any of the seven gates is incomplete, a strong judgment cannot be maintained.
  8. Form at least two mechanism candidates; unless the evidence is sufficient to show that there is only one mechanism.
  9. Fully absorb the concepts that bear the judgment function: read the corresponding concept card, and use worksheets/concept-fidelity-check.md to perform the fidelity check.
  10. Use worksheets/source-continuity-check.md to check whether only isolated concept cards were read, whether v5 adjacent constraints were missed, or whether a downgrade is needed.
  11. Use worksheets/source-anchor-integrity-check.md to check whether the central proposition, mechanism candidates, high-risk concepts, and action boundaries can back-refer to the capsule source anchors; content that cannot back-refer can only be labeled “inference in this text / expressive translation / external thought mapping” and must not be written as the original meaning of CrossFrame v5.
  12. Decide the judgment tier: lightweight observation, open assertion, full diagnosis, strong judgment, low-condition exploratory action, exit transfer. If joint reading is required but not performed, or the source anchors are insufficient, a strong judgment cannot be maintained.
  13. First output a visible reasoning outline, then choose a template output: first speak in reality language, then attach internal mapping as needed.

Reading rules

  • By default comply with references/runtime-read-policy.md: do not read evals/、examples/, complete success cases, complete failure cases, or the full v5 large index. They are only used for development stress testing, regression validation, style debugging, explicit user requests for source audit, or targeted supplementary reading after source anchor failure.
  • Ordinary diagnosis: read protocols/diagnosis-protocol.md, and use worksheets/intake-worksheet.md、worksheets/seven-gates-worksheet.md、worksheets/evidence-ledger.md、worksheets/mechanism-candidates.md。
  • deduction, subsequent trajectory, path expansion, branch endpoints: read protocols/inference-protocol.md and templates/inference-output.md, and append state coordinates, long-term evolution, and governance continuity packs as needed.
  • Judgments with low to moderate confidence: read protocols/open-assertion-protocol.md、worksheets/open-assertion-record.md、templates/open-assertion-output.md and v5-open-assertion-proposition-pack。
  • Questions involving high responsibility, strong power density, sanctions, reputation, rights, resources, or public memory: read protocols/anti-capture-protocol.md、worksheets/high-responsibility-check.md, and append v5-low-power-protection-pack、v5-evidence-downgrade-action-ceiling-pack。
  • Strong judgments affecting eligibility, reputation, resources, rights, disposal, or public memory: read protocols/proposition-verification-protocol.md、worksheets/proposition-verification.md、worksheets/prospective-registration.md、templates/strong-judgment-output.md, and append v5-strong-judgment-eight-pack。
  • AI reports, compliance materials, polished reports, institutional self-assessments, model diagnoses: read v5-ai-process-artifact-boundary-pack; it must be declared that processual artifacts must not serve as proof of reality.
  • Objects that will change behavior, identity, evidence, or boundaries because they are observed, named, made public, or handled: read protocols/high-reflexivity-protocol.md、worksheets/reflexivity-state-transfer.md、templates/high-reflexivity-output.md and v5-observation-reflexivity-release-pack。
  • Scenarios of intimate relationships, family, friends, caregiving, unilateral承接, interpretive labor, and love being demanded: read protocols/intimate-relationship-protocol.md、worksheets/intimate-relationship-light-check.md、templates/intimate-relationship-output.md, and first read v5-love-trapped-trauma-pack and v5-low-power-protection-pack。
  • Scenarios of system stagnation, trauma, repair, exit transfer, and reconstruction: read protocols/healing-transfer-protocol.md、worksheets/healing-transfer-map.md、templates/healing-transfer-output.md and v5-action-healing-transfer-pack。
  • Public institutions, platform governance, public commitments, and public issues with high power density: read protocols/public-institution-protocol.md、worksheets/public-institution-check.md、templates/public-institution-output.md, and append v5-public-power-institution-pack、v5-evidence-downgrade-action-ceiling-pack、v5-low-power-protection-pack。
  • When CrossFrame may be treated as a universal theory, a domain substitute, a personality judgment tool, or endorsement material for AI compliance: read protocols/framework-boundary-protocol.md、worksheets/framework-boundary-check.md、references/framework-ontology-protection.md and v5-use-boundary-governance-pack。
  • Long-term evolution, stage assessment, organizational/relationship/institutional cyclical change: read protocols/lifecycle-diagnosis-protocol.md、worksheets/lifecycle-stage-record.md、templates/lifecycle-output.md and v5-state-coordinate-lifecycle-pack. Stages 0-6 can only serve as local state coordinates; writing them as linear destiny is prohibited.
  • Strategic advancement, long-term repair, sub-anchor closed loop, why busy but without accumulation: read protocols/progression-protocol.md、worksheets/sub-anchor-progression.md、templates/progression-output.md and v5-long-evolution-progression-field-pack。
  • Positive and negative potential fields, sedimentary base, autonomous dissociation, protective exit: read protocols/field-dissociation-protocol.md、worksheets/field-dissociation-check.md and v5-long-evolution-progression-field-pack。
  • Regulation, early warning, solvency constraints, polycentric governance, generation of receivers, and intergenerational succession: read protocols/governance-continuity-protocol.md、worksheets/governance-continuity-check.md、templates/governance-continuity-output.md and v5-governance-continuity-multicenter-pack。
  • Civilizational scale, historical scale, ultra-large-scale circles, or grand public judgments: read protocols/large-scale-stress-test-protocol.md、worksheets/large-scale-stress-test.md、templates/large-scale-stress-output.md, and first downgrade-check the evidence and release gates.
  • For ordinary people, management, institutional public affairs, technology governance, or other AI software rewriting expressions: read protocols/expression-translation-protocol.md、references/expression-translation-table.md、templates/expression-translation-output.md and v5-domain-translation-normative-source-pack。
  • Concept explanation, concept boundaries, and idea-explanation questions: read protocols/concept-explanation-protocol.md、references/concepts-minimal-set.md、references/v5-term-fidelity.md、v5-core-concept-integrity-pack, then read the necessary concept cards as needed.
  • Abstract questions such as philosophy, meaning, first cause, what life is, nihilism, and reasons for existence: prioritize the concept explanation protocol, first perform scale decomposition and structural open assertions; only when it cannot be converted into any structural question should you fall back to protocols/framework-boundary-protocol.md。
  • If the final output is to use high-risk concepts such as 承接/回流, 开放断言, 尺度转移, 观测反身性, 权力封闭, 低条件试探行动, 爱/开放行动, 主体/责任链, 证据成本, 机制候选, 判断档位, 退出转移, or 修复副产品, you must first read the corresponding concept cards and the v5 continuous reading pack; you cannot make fine-grained judgments based only on the minimal concept set.

Output Rules

  • Default output should be short and clear. Unless the user explicitly requests an extremely minimal conclusion, first present a “reasoning outline”; do not display the complete worksheet.
  • The reasoning outline must include: diagnostic object, factual boundaries, scale window, seven-gate review, mechanism candidates, judgment tier, the concepts read this time or fidelity checks, this time's v5 continuous linked-reading pack, reading-state capsule summary, and next-step observation or action.
  • In deep, audit, high-responsibility, public-institution, intimate-relationship, long-term-evolution, and article-output scenarios, the reasoning outline must display “this time's continuous linked-reading pack”; for ordinary lightweight questions, it may state “not triggered.”
  • The reasoning outline may only contain an outline, not lengthy internal reasoning; it is used to let the user see the reasoning path and also to constrain subsequent output from skipping steps.
  • Only when the user requests “complete reasoning process,” “internal mapping,” “worksheet,” or “audit” should the complete worksheet be expanded.
  • By default, speak plainly first and do not pile up terminology. The first paragraph must let people who have never read the framework understand “what happened, why it is stuck, and what to look at next.”
  • Always distinguish: source, fact, evidence, explanation, mechanism candidate, judgment tier, and action ceiling.
  • Terminology may only serve as an additional mapping and must not be treated as the conclusion itself. Do not replace reasoning with “this is a typical X, therefore Y.”
  • Before output, it must pass the expression gate: after deleting all framework terminology, the core judgment can still be understood by an ordinary user.
  • Structural diagnosis must not be turned into personality judgment, fate prophecy, ideological labeling, or moral authorization.
  • Scale elevation must not be used to erase low-scale pain, pressure, dereliction of duty, and chains of responsibility.
  • "Love" must not be described as a command, a justification, or a unilateral demand for endurance.
  • AI-generated compliance materials, polished reports, and self-assessment texts must not be treated as high-cost evidence.
  • Strong judgment must not be used to bypass proposition verification; open assertions cannot serve as the basis for high-responsibility handling.
  • High-reflexivity objects must not be recursively expanded without limit; after the third layer, if there is no new high-cost evidence or structural variable, it must be closed or downgraded.
  • In intimate relationship scenarios, first protect pain, safety, and boundaries, and do not press the responsibility for repair back onto the injured party.
  • Healing and transfer only provide structural action boundaries and do not replace medical, psychological, legal, safety, or organizational handling.
  • If evidence is insufficient but the problem is urgent, output low-risk, retractable, observable small actions instead of pretending that a strong diagnosis has already been completed.

Expression gate

Before final output, internally check four questions:

  1. Can the first paragraph explain the problem clearly without terminology?
  2. Can the user know which facts this judgment comes from, rather than from conceptual application?
  3. Have terms such as "承接、回流、尺度、开放断言" been translated into real-world behaviors?
  4. Has an observable signal or action boundary been given?

If any one does not pass, rewrite the expression first, then output.

Reasoning outline

By default, place a brief outline before the output:

  • Diagnostic object:
  • Fact boundary:
  • Scale window:
  • Seven-gate review:
  • Mechanism candidates:
  • Judgment tier:
  • Concepts read this time:
  • This v5 continuous reading package:
  • Next step:

This outline is not a complete worksheet, nor is it a lengthy chain of reasoning. Its purpose is to let the user see that this output indeed first defines the object, checks the evidence, compares mechanisms, and then gives a judgment.

Core materials

  • references/runtime-read-policy.md: The lightweight reading strategy under normal operation, controlling the default non-reading boundary for eval/examples, complete cases, and large source modules.
  • references/continuity-closure-map.md: The lightweight runtime graph of the v5 continuous reading package closure.
  • references/v5-source-spine.md: The v5.0 original text title hierarchy, chapter order, paragraph ranges, adjacency relations, table index, and default continuous reading package.
  • references/v5-section-digest-index.md: The v5.0 section-by-section faithful summaries, boundaries that must not be misread, and adjacent co-reading reminders.
  • references/v5-coverage-map.md: The coverage map from v5.0 chapters to skill modules, protocols, worksheets, and continuous reading packages.
  • references/v5-term-fidelity.md: The v5.0 terminology fidelity table to prevent compression distortion.
  • references/v5-material-selection-map.md: The selection map for v5.0 source modules, continuous reading packages, protocols, and templates.
  • references/continuity-bundles.md: v5.0 continuous co-reading package index.
  • references/continuity-bundles/v5/: 26 v5 standalone co-reading packages.
  • references/read-routing-map.md: Select protocols, worksheets, concept cards, and templates by request type.
  • templates/read-state-capsule.md: Capsule template for this v5 source modules, entry package, mandatory co-reading closure, source anchors, and downstream reading strategy.
  • references/crossframe-v2-core.md、references/v2-*、references/v3-*: Historical baseline, read only during version tracking or rollback audit.
  • references/concepts-minimal-set.md: Minimal concept set.
  • references/framework-ontology-protection.md: Framework ontology protection, anti-domain colonization, anti-model colonization, and concept change rules.
  • references/guardrails.md: Anti-misuse rules.
  • references/diagnostic-dimensions.md、references/diagnostic-toolbox-index.md: Read complex cases on demand.
  • references/theory-backend-index.md: Deep theory index for root assumptions, core inferences, full-cycle evolution, progressive patterns, polycentric governance, etc.
  • references/expression-translation-table.md: Translate backstage concepts into ordinary people, management, institutional, and technical governance contexts.
  • references/concept-cards/: High-risk concept cards.
  • worksheets/seven-gates-worksheet.md: Seven-gate review table.
  • worksheets/source-continuity-check.md: Before output, check whether too little was read, whether passages were taken out of context, or whether continuous constraints from the original text were missed.
  • worksheets/source-anchor-integrity-check.md: Before output, check whether the central proposition, mechanism candidates, concepts, action boundaries, and article translation can refer back to the capsule source anchors.

High-risk concept gate

The following concepts cannot be understood only literally; once they carry a judgment role, the corresponding concept cards and v5 co-reading packages must be read:

  • Inheritance / reflow: read references/concept-cards/chengjie-huiliu.md, and co-read v5-core-concept-integrity-pack。
  • Open assertion: read references/concept-cards/open-assertion.md, and co-read v5-open-assertion-proposition-pack。
  • Scale shift / scale elevation: read references/concept-cards/scale-transfer.md, and co-read v5-cross-scale-context-translation-pack。
  • Observational reflexivity: read references/concept-cards/reflexivity.md, and co-read v5-observation-reflexivity-release-pack。
  • Power closure / anti-capture: read references/concept-cards/power-closure.md, and co-read v5-public-power-institution-pack and v5-low-power-protection-pack。
  • Low-condition exploratory action: read references/concept-cards/low-condition-action.md, and co-read v5-diagnosis-admission-downgrade-exit-pack。
  • Love / open action / do not waste love: read references/concept-cards/love-open-action.md, and co-read v5-love-trapped-trauma-pack。
  • Subject / chain of responsibility: read references/concept-cards/responsibility-chain.md, and co-read v5-responsibility-intervention-separation-pack。
  • Evidence cost / weak signals / AI compliance materials: read references/concept-cards/evidence-cost.md, and co-read v5-source-evidence-separation-pack and v5-ai-process-artifact-boundary-pack。
  • Mechanism candidates: read references/concept-cards/mechanism-candidates.md, and pass the seven gates.
  • Judgment tier: read references/concept-cards/judgment-grades.md, and co-read v5-evidence-downgrade-action-ceiling-pack。
  • Exit shift: read references/concept-cards/exit-transfer.md, and co-read v5-action-healing-transfer-pack。
  • Repair byproduct / pseudo-repair: read references/concept-cards/repair-byproduct.md, and co-read v5-action-healing-transfer-pack。
  • Lifecycle / stages: read protocols/lifecycle-diagnosis-protocol.md, and co-read v5-state-coordinate-lifecycle-pack。
  • Framework governance / falsification / benign demise: read references/concept-cards/framework-governance-falsification.md, and co-read v5-framework-self-diagnosis-falsification-pack。
  • Unable to exit the subject / complex trauma / no healthy baseline: read references/concept-cards/trapped-subject-trauma-baseline.md, and co-read v5-love-trapped-trauma-pack。
  • Metaphor drift / source transparency / normative premises: read references/concept-cards/metaphor-source-transparency.md, and co-read v5-domain-translation-normative-source-pack。
  • Usage threshold debt / instrumentalization / split protocol: read references/concept-cards/accessibility-toolization-split.md, and co-read v5-toolization-accessibility-release-pack。
  • Observation Convergence / Entropy Increase Boundary: read references/concept-cards/observation-entropy-contraction.md, and jointly read v5-observation-reflexivity-release-pack。

minimum qualifying standard

A qualifying CrossFrame output must be able to answer:

  • What object are we actually diagnosing?
  • Which are sources, which are facts, and which are merely interpretations?
  • Which scale window are we currently in?
  • Which of the seven gates passes, and which gate causes downgrading?
  • What are at least some mechanism candidates?
  • Is proposition verification, the Strong Judgment Eight-Piece Set, high-reflexivity handling, the intimate relationship lightweight entry point, healing transfer, or a public institution special project needed?
  • Who bears the cost, and who has the conditions to change?
  • Which high-risk concepts does this judgment depend on, and have the complete concept cards and the v5 joint reading pack been read?
  • By what evidence can this judgment be retracted?
  • Is the next step observation, repair, exploratory action, or exit transfer?
  • Is the lifecycle judgment written as local-state coordinates rather than linear destiny?
  • Does this session trigger the v5 continuous joint reading pack, and does it avoid reading only isolated concept cards?
  • Is it generated v5-read-state-capsule, and does it make the central proposition, mechanism candidates, high-risk concepts, and action boundaries point back to the source anchors?