dingxingdi
- 389 skills
- 0 followers
- 20 hours ago last updated
- ▌ Domain Terminology Translation · dingxingdi bundleUse this skill when the user wants translation data for specialized professional fields like medicine, law, science, mathematics, or academic research, where words have specific technical meanings or require formal scholarly syntax. Trigger it for requests like 'translate this as a research abstract,' 'maintain the mathematical logic exactly,' 'ensure the word choice matches the scientific context,' or 'translate this legal, medical, or logical paper without simplifying the wording or correcting its intentional academic flaws.'
- ▌ Text Image Machine Translation · dingxingdi bundleUse this skill when the user wants to translate text found inside an image (like a photo of a sign, a screenshot, or a damaged manuscript) OR when the user provides noisy text that must be translated by referencing an accompanying image. Trigger it for layman requests like 'translate the text in this picture,' 'decode the ancient script on this stone,' 'read the words in this screenshot,' or 'fix the typos in this description using the picture to ensure the colors match what is actually shown.'
- ▌ Length Requirement Adherence · dingxingdi bundleUse this skill when the user cares about how short or how long the writing should be. Trigger it for requests like “keep it under 500 words,” “make it at least 25 sentences,” “give me 3 paragraphs,” or “write a 4,000-word article.” It is especially useful when the agent must hit a target length without padding, truncation, or collapse in quality.
- ▌
- ▌ Table Text Numerical Reasoning QA · dingxingdi bundleUse this skill when the user wants the agent to perform document-grounded math, execute multi-step financial formulas, or perform 'intrinsic fact-checking' to recover masked or missing numerical values from tables. Trigger it for requests like 'calculate the ROI', 'verify if this revenue figure is consistent with the balance sheet', 'find the year-over-year growth rate', 'recover the missing value in this report', or 'perform latent variable inference to find the equity investment'. This skill is essential for detecting numerical hallucinations and ensuring calculated values are strictly faithful to the provided document context.
- ▌ Multi Turn Preference Evaluation · dingxingdi bundleUse this skill when a user wants evaluator data to check how well an assistant handles context over time, ranging from short-term conversational follow-ups to long-term multi-session behavioral adaptation. Trigger it when people say things like 'compare these two chats', 'check if the bot remembered my follow-up', 'see if it inferred my unstated needs from last week', or 'judge which bot handles the second part better'.
- ▌ Bulk Iterative State Coordination · dingxingdi bundleUse this skill when the user wants to process a long list, a large table, or multiple items one by one. Trigger it for requests like “collect all the names from the table,” “copy every row from the list,” “loop through the folders without missing any,” “make sure you get all 50 items across the pages,” or “process these invoices skipping the ones I've already done.” This skill focuses on tracking progress during repetitive loops to ensure no items are omitted and no actions are redundant.
- ▌ Dexterous Spatial UI Manipulation · dingxingdi bundleUse this skill when the user wants tasks requiring precise physical interaction across coordinates, such as drag-and-drop, selecting specific text spans, adjusting sliders, or drawing/arranging elements based on spatial constraints. Trigger it for requests like “highlight the second sentence,” “select the paragraph starting with 'For',” “move the file to the folder,” “set the slider to 75%,” “draw a circle inside the square,” or “draw a 3x3 grid.” This skill is for GUI tasks where the agent must resolve spatial start/end points, compute geometric bounding boxes, or maintain continuous motion control to manipulate UI elements precisely.
- ▌ Conflict Aware Faithful Grounding · dingxingdi bundleUse this skill when the user wants to synthesize data testing an agent's ability to handle contradictions, misleading facts, factual errors, or semantic drift (e.g., from AI-generated summaries). Trigger it for requests like 'make the documents disagree with each other', 'test if the model spots self-contradicting files', 'see if it gets tricked by summarized text', 'check for mistakes in rewritten news', or 'see if the agent follows a lie in the text over what it knows is true'. This skill covers factuality detection, self-contradiction, and resolving information fusion or omission.
- ▌ Embedded Instruction Resistance · dingxingdi bundleUse this skill when the user wants translation data where the source text looks like an instruction and may tempt the model to do the wrong job, such as solving a math problem, answering a question, or writing code instead of translating. Trigger it for requests like "make sure it translates the problem instead of solving it," "the input itself looks like a command," or "test whether it can ignore fake instructions inside the text being translated."
- ▌ Action And Motion Understanding · dingxingdi bundleUse this skill when the user asks 'how many times did [action] happen?', 'what is the player doing?', 'where did the ball go?', 'is this a complex move?', 'what happens next?', or 'compare these two athletes'. It is triggered for tasks assessing high-frequency sports dynamics, counting rapid repetitive actions (like rally counts), identifying specific kinetic stroke techniques, predicting immediate athletic outcomes, and ranking technical difficulty based on observed action units.
- ▌ Entity And Key Detail Retrieval · dingxingdi bundleUse this skill when the user asks 'where are my keys?', 'what specific tool did they use?', 'find the object of interest among the mess', 'what does the sign in the background say?', or 'extract the numbers/names so I can look them up'. Trigger it for tasks extracting exact metrics, reading subtle OCR/text from the environment, identifying fleeting details, and performing object-centric reasoning where a specific entity or variable must be isolated from distractors and complex scenes.
- ▌
- ▌
- ▌
- ▌
- ▌
- ▌ Code Mixed Multilingual Handling · dingxingdi bundleUse this skill when translating content that mixes two or more languages (like Hinglish, Spanglish, or Manglish) into a single language, or vice versa. It is especially important for handling casual dialogues where people switch languages mid-sentence, such as in chat transcripts or social media. Trigger this for requests like 'translate this mixed English-Mandarin chat,' 'put this Hinglish text into pure English,' 'make sure you don't ignore the non-English words,' or 'test if the translator understands how bilingual speakers talk.'
- ▌ Lexical Substitution Translation · dingxingdi bundleUse this skill when the user wants the translator to handle everyday wording changes like "turn this into French," "render this in Arabic," or "put this into Spanish" instead of a canonical "translate" instruction. Trigger it when the user says things like "people ask in different words," "use casual phrasing instead of textbook phrasing," or "make sure the system still gets the job when the wording changes but the meaning stays the same."
- ▌ Citation Grounded Report Synthesis · dingxingdi bundleUse this skill to generate high-quality, professional investigative reports that mirror expert human standards and satisfy the Deep Reasoning (Depth) and Wide Aggregation (Width) dimensions of research. Trigger this skill when the user asks for 'analyst-level research,' 'deep-dive investigative reports,' or when they need to 'verify findings against expert standards' on the live web. It is especially useful for requests like 'do some deep research into this,' 'make sure the analysis goes beyond just summarizing facts,' 'trace the specific history across the web,' or 'reconcile complex data into a structured case study.' This skill ensures the output provides both logical deduction to identify hidden entities and comprehensive attribute gathering to ensure data completeness.
- ▌ Cross Lingual Ecosystem Navigation · dingxingdi bundleUse this skill when the agent needs to move beyond English-centric web pages to search, reason, or perform interactive tasks (like shopping or attribute selection) on non-English websites. Trigger it for requests like “find a product on a French site,” “use the Chinese version of this shop,” “buy a yellow coat from the local store,” or “check these details across different language portals.” It is designed for scenarios where the challenge is not just translation, but navigating UI elements, buttons, and menus that are labeled in a foreign language.
- ▌ Evidence Triage And Reconciliation · dingxingdi bundleUse this skill to verify the truthfulness of web-based claims, handle conflicting reports, and self-correct agent failures by decomposing complex answers into verifiable sub-questions. Trigger it for requests like “is this statement true?”, “fact-check this,” “verify the findings,” or “verify if the previous answer makes sense.” It is essential for avoiding 'premature conclusions' and 'hallucinated claims' by leveraging verification asymmetry—breaking a major claim into atomic 'Yes/No' sub-tasks that are easier to ground in primary documentation than the initial open-ended query.
- ▌ Element Complete Summarization · dingxingdi bundleUse this skill when the user wants a summary that captures the core facts cleanly instead of giving a vague gist. Trigger it for requests like “summarize the article but keep who/when/what happened/what resulted,” “don’t miss the key outcome,” or “compress the news without dropping important details.” It is especially useful when summaries often sound smooth but leave out the most important factual element.
- ▌
- ▌ Objective Correctness Adjudication · dingxingdi bundleUse this skill when verifying if an answer is factually, mathematically, or logically correct against references, gold-standard contexts, verifiers, or massive source texts. Trigger it when users ask to 'check if the bot used the right formula', 'verify the math logic', 'find if the assistant hallucinated a detail', 'verify factuality in this 100-page document', or 'verify if the Telugu math solution is actually correct'.
- ▌ Graph Relational Subgraph Reasoning · dingxingdi bundleUse this skill when you need to navigate structured knowledge bases, relational triplets, or knowledge graphs to connect different facts. It is triggered by requests like 'connect the dots between these news updates', 'find the person who meets all these shared conditions', 'what are all the projects this artist is linked to', or 'who is the owner of the company that sold the most units'. This skill emphasizes traversing explicit symbolic paths (like Subject-Predicate-Object) to find intersections or multi-step connections that a single search cannot reveal.
- ▌ Gender Fair Inclusive Translation · dingxingdi bundleUse this skill when the user wants translation data that avoids gendered biases, such as making sure people aren't just referred to as 'men' by default. Trigger it for requests like 'make the translation gender-neutral,' 'use inclusive Italian,' 'don't assume everyone is male in the translation,' or 'use the schwa or asterisk for gender-neutral endings.' It covers strategies like using collective nouns, providing both male and female forms, or using modern non-binary symbols.
- ▌ Semantic Perturbation Translation · dingxingdi bundleUse this skill when the user wants translation data with bigger but still meaning-preserving prompt changes, such as role prompts, extra setup language, or semantically equivalent rewrites that sound very different on the surface. Trigger it for requests like "wrap the ask in a role," "change the wording a lot but keep the job the same," or "see if it still translates when the instruction is semantically the same but phrased very differently."
- ▌ Sentential Rephrasing Translation · dingxingdi bundleUse this skill when the user wants translation data where the ask is phrased in an unusual sentence shape, such as an indirect request, a longer setup sentence, or a differently ordered instruction. Trigger it for requests like "say the same thing but in a more roundabout way," "people do not always ask in one short command," or "test whether the system still translates when the request is worded awkwardly but means the same thing."
- ▌ Simultaneous Adaptive Translation · dingxingdi bundleUse this skill when the user wants to translate content in real-time or as a stream, where the translation must start before the source text is finished. It is triggered by layman requests like 'translate this live,' 'start translating as I speak,' 'don't wait for the whole sentence to finish,' or 'real-time subtitles for this stream.' It allows for adaptive reading and writing based on specific latency needs (low, medium, or high).
- ▌ Multi Turn Revision Consistency · dingxingdi bundleUse this skill when a user provides follow-up feedback to a previously generated document, such as 'shorten the intro,' 'now add a section on pricing,' or 'tweak the formatting to use bullets.' It is triggered when writing is an iterative dialogue where the agent must integrate new instructions without 'breaking' or regressing on the content, citations, or presentation quality already established in earlier turns.
- ▌ Experiment Driven Model Improvement · dingxingdi bundleUse this skill when the user wants data where the agent must improve an existing ML training setup or develop novel research methodologies by editing code and running experiments. Trigger it for requests like 'generate ML debugging and improvement tasks', 'make train.py optimization problems', 'create model tuning workflows', 'develop a new LLM merging technique', 'orchestrate end-to-end ML research', or 'propose a novel ML research baseline'. Do not use it for broad open-ended Kaggle competition workflows or strictly paper-reproduction tasks from scratch without an iterative improvement objective.
- ▌
- ▌
- ▌ Temporal Ordering And Localization · dingxingdi bundleUse this skill when the user asks 'when exactly did the event happen?', 'spot the shot change', 'why did the camera zoom into the face?', 'is this a close-up or a full-body view?', or 'find the boundary where the speaker starts walking'. It is triggered for tasks assessing precise temporal grounding, identifying cinematographic shot scales (like Close-up, Medium, or Full Shot), and parsing the logical relationship between audio/emotional shifts and structural camera transitions across a multi-shot video timeline.
- ▌ Stateful Multi File Code Refactoring · dingxingdi bundleUse this skill when the user wants to reorganize, clean up, or restructure code across multiple files without changing its functional behavior. Trigger it for requests like 'refactor this project', 'clean up the code structure', 'move these functions to a separate utility file', 'consolidate these classes', or 'change the naming convention throughout the whole repository'. It is specifically designed for complex, multi-step changes where the agent must track its own previous edits and handle temporary errors while moving logic between different files.
- ▌
- ▌ Citation Grounded Factuality Checking · dingxingdi bundleUse this skill when the user wants the agent to judge whether an answer/document is securely backed by authentic evidence, verify exact pinpoint highlight spans, validate strict antecedent document grammar (e.g. tracking 'the' to 'a'), or audit sites for safety/phishing risks and machine-generated stylistic anomalies. Trigger for 'verify this highlighted claim,' 'check if this legal clause has grammatical consistency,' 'is this link a scam?', or 'identify if any of these sentences were polished by AI.' It is essential for eliminating hallucinations, detecting structural phrasing errors, and ensuring document authenticity.
- ▌
- ▌ Bias Resistant Human Aligned Judging · dingxingdi bundleUse this skill when evaluating if an AI judge exhibits stable, objective, and logically sound reasoning, particularly against situational and statistical biases. Trigger it when requests mention 'test if the judge is tricked by a sob story', 'check if the evaluator is swayed by flattery', 'verify if the model is too sure too early based on vague clues', 'check if confidence increases just because the chat got longer', or 'test if the judge is just copying the reference answer'. It is designed for meta-evaluation, detecting susceptibility to rhetorical persuasion, 'Length Artifacts' (monotonicity bias), Criteria Entanglement (halo effect), and 'Solution Fixation' induced by references.
- ▌ Grounded Task Infeasibility Detection · dingxingdi bundleUse this skill when the user asks for a task that might be impossible, broken, or based on features that do not exist. Trigger it for requests like “tell me if you can't do this,” “explain why the button is missing,” “don't just fail, show me why it’s not working,” or “check if this command is actually possible in the current app version.” This skill ensures the agent identifies deprecated features, hallucinated commands, or environmental blockers and provides a logical reason for stopped execution rather than a generic failure.
- ▌ Semantic Gui Hierarchical Exploration · dingxingdi bundleUse this skill when the user wants to find a button or setting that isn't on the front page and requires digging through menus. Trigger it for requests like “I can't find where to change my avatar,” “where is the customer service hidden,” “find the setting to turn off auto-play,” or “explore the app to find the draft folder.” This skill focuses on searching through deep app hierarchies and figuring out which icons or text labels to click when the path to the goal is not obvious or mentioned in the prompt.
- ▌
- ▌
- ▌
- ▌ Implicit Commonsense Consistency · dingxingdi bundleUse this when the user wants planning data about plans that should 'make sense' or 'be safe' even if the user did not spell out every rule. Trigger it for requests like 'make itinerary tasks where the plan should feel realistic', 'make sure the agent doesn't leave the stove on', 'create chores where the robot must avoid making the floor slippery', or 'create questions where the agent must avoid repeating the same attraction or inventing fake flights.'
- ▌ Resource Lifecycle Orchestration · dingxingdi bundleUse this when the user wants data tracking procedural cycles, tool usages, or tracking task progression using history logs or videos. Trigger it for requests like 'what is the next move for this recipe', 'make maintenance tasks with strict cleaning orders', 'look at my progress video and tell me what to do now', or 'create plans where forgetting to reset a tool ruins the workflow.'
- ▌ State And Attribute Change Tracking · dingxingdi bundleUse this skill when you need to know 'if the video looks real', 'if the background is glitching', 'is the motion smooth or choppy?', 'does the person change their appearance mid-video?', or 'how much movement is actually happening?'. It is essential for auditing visual quality, identifying generative AI artifacts like warping backgrounds or flickering subjects, and checking if the dynamic motion matches the expected intensity. Trigger it whenever the user wants a quality check on how natural or consistent a video is, especially for identifying fakes or poorly generated content.
- ▌ Long Form Breadth Depth Coherence · dingxingdi bundleUse this skill when the user wants a long report, article, monograph, or multi-thousand-word draft that still needs to stay organized and substantive. Trigger it for requests like “write a 4,000-word report,” “draft a 10,000-word article,” or “expand this into a full-length overview without repeating yourself.” It is especially useful when long outputs tend to become shallow, repetitive, or incoherent.
- ▌ Paper To Code Implementation Fidelity · dingxingdi bundleUse this skill when the user wants to implement complex scientific algorithms or methodologies from research papers, LaTeX formulas, or technical documentation into a code repository. Trigger it for requests like 'implement the research method from this PDF', 'code the formula from Section 3', 'refine my AI model implementation using these expert reviewer notes', or 'fix the discrepancies between our math derivation and the current code'. It is specifically designed for high-precision implementations requiring deep grounding in research artifacts, domain-specific terminology, and iterative refinement of mathematical logic based on hierarchical expert feedback.
- ▌ Policy Aligned Repository Stewardship · dingxingdi bundleUse this skill when you want to evaluate an agent's commitment to safety policies and engineering integrity, especially when task completion is hindered by operational pressure or tempting shortcuts. It is triggered by requests such as 'test if the agent uses harmful ways to meet a deadline', 'simulate resource scarcity to see if the agent bypasses security', 'check for propensity to misuse powerful tools under stress', or 'ensure the agent doesn't take misaligned shortcuts to achieve a goal'. It is crucial for detecting when an agent prioritizes instrumental efficiency over ethical constraints or prescribed safe procedures.
- ▌
- ▌
- ▌
- ▌ Global Dag Topology Planning · dingxingdi bundleUse this skill when the user wants the agent to create a complete map, blueprint, or flowchart of actions before doing anything. It should be triggered by layman phrases like 'plan the whole process at once,' 'show me the arrows between the tools,' 'don't just guess step-by-step,' 'make a flowchart of the dependencies,' or 'give me a global plan for this complex request.' This is specifically for situations where tasks have nested logic or multiple parallel paths that need to be understood as a Directed Acyclic Graph (DAG) to avoid getting stuck in local loops or missing the big picture.
- ▌ Long Horizon Itinerary Sequencing · dingxingdi bundleUse this when the user wants high-resolution travel plans spanning multiple days or cities with precise hour-by-hour schedules, meal gap logic, and clustering based on public transit proximity. Trigger it for requests like 'make a detailed trip schedule with timestamps', 'give me a multi-day plan that includes transit stop distances', 'create an itinerary with specific meal times and durations', or 'build a travel route that matches my adventurous or laid-back persona preferences.'
- ▌ Morphosyntactic And Lexical Fidelity · dingxingdi bundleUse this skill when translating content into languages where words change their form through prefixes or suffixes to show grammar, or where the same word means different things. Trigger it for requests like 'ensure the prefix matches the pronoun,' 'don't mess up the verb endings in agglutinating languages like Ibibio,' 'keep the grammar straight in complex sentences,' or 'make sure words with many meanings are translated correctly based on the context.'
- ▌ Transformationally Robust Patch Repair · dingxingdi bundleUse this skill when the user wants bug-fix or security-fix data that stays the same in meaning but looks different on the surface, so the agent cannot rely on memorized code patterns. Trigger it for requests like 'make decontaminated repair tasks', 'generate bug fixes with renamed or refactored code', 'create harder unseen patching data', or 'give me semantically same but structurally different repair problems'. Do not use it for standard repair data where surface form can directly reveal the answer.
- ▌
- ▌ Multi Document Integrative Reasoning QA · dingxingdi bundleUse this skill when the user wants questions that require connecting information across several files, navigating multi-page documents, or resolving facts across documents in different languages. It handles organizational information gathering where data is siloed across multiple repositories or people, requiring the navigation of 'redirections' (knowing who knows) and the reconstitution of 'split documents' (data distributed across separate sources). Trigger it for requests like 'combine details from these reports,' 'reason across this multi-hop chain,' 'follow the trail between these files,' 'find the connection between the person mentioned here and the data in that file,' or 'reach out to different sources to compile a complete answer.'
- ▌ Agent Trajectory Safety Risk Diagnosis · dingxingdi bundleSkill: agent-trajectory-safety-risk-diagnosis
- ▌ Audio And Speech Response Adjudication · dingxingdi bundleSkill: audio-and-speech-response-adjudication
- ▌ Live Web Multimodal Information Seeking · dingxingdi bundleUse this skill when the user wants to extract specific factual data that is deeply embedded within complex websites or applications, requiring multi-step stateful interaction. Trigger it for requests like “find the HS code for this product,” “verify the business license on the official registry,” “lookup the hidden ECCN classification,” “check the merchant’s WHOIS details,” or “scan this site for any hidden gambling or phishing content.” This skill is for GUI tasks where the answer cannot be found in surface-level search snippets and requires navigating through dynamically loaded subpages, tabs, and interactive elements to retrieve high-stakes intelligence.
- ▌ Nested Aggregation And Filter Reasoning · dingxingdi bundleSkill: nested aggregation and filter reasoning
- ▌ Cross Lingual Instruction Consistency · dingxingdi bundleUse this skill when the user provides translation tasks where the instruction, the source text, and the target output involve different languages (n x n configurations). Trigger it for layman requests like 'English prompt for Spanish text to Japanese,' 'mismatched languages between the ask and the content,' or 'test if the translator gets confused when the command is in English but the text is in German and the answer should be in Dutch.'
- ▌ Target Style And Persona Adaptation · dingxingdi bundleUse this skill when you need the writing to mimic a specific famous author, reflect a unique persona, neutralize subjective bias into an encyclopedic tone, or copy the exact structural pacing of a professional genre. Trigger it for requests like 'write this in the voice of Jane Austen,' 'adopt the vocabulary of a cynical detective,' 'make this sound like a formal legal brief,' 'remove the bias and make it sound like a Wikipedia article,' or 'imitate the structure of a TED talk.' It is essential for capturing unconscious linguistic fingerprints, domain-specific collocations, and thematic structural beats while resisting generic AI stereotypes.
- ▌ Competitive Algorithmic Problem Solving · dingxingdi bundleUse this skill when the user wants data for solving elite-level algorithmic puzzles (like NOI or ICPC) that require 'aha' moments, advanced mathematical insights, and performance-critical implementations. Trigger it for requests like 'generate NOI-level coding challenges', 'make tasks that need efficient C++ implementations for math puzzles', 'create problems where the agent must refine code based on TLE or MLE feedback', or 'give me high-difficulty competitive programming trajectories'. This skill is specifically for scenarios where naive solutions fail and the agent must utilize an execution feedback loop to reach an AC (Accepted) state under tight time and memory limits.
- ▌ Subjective Qualitative Response Scoring · dingxingdi bundleUse this skill when you need to score single open-ended answers on subjective qualitative dimensions like tone, coherence, formatting adherence, cultural politeness, or complex role-play behavior. Trigger it when people say 'score how well the answer flows', 'check if the response is concise enough', 'evaluate if the doctor sounds professional', 'judge if the bot correctly plays the villain persona', or 'tell me if the Japanese answer uses the proper polite honorifics'.
- ▌ Unified Content Moderation Adjudication · dingxingdi bundleUse this skill when an evaluation agent needs to moderate content, detect hate speech, identify offensive language, or spot implicit bias in human-ai interactions. Trigger it when users use casual language like 'clean up this chat', 'check for mean comments', 'is this post offensive?', 'spot the hidden racism', or 'scan for hateful slang across different cultures'.
- ▌ Gui Self Reflection And Error Correction · dingxingdi bundleUse this skill when the agent needs to handle environmental obstacles like pop-up advertisements blocking the target, or when it must correct common interactive widget errors such as pre-filled text in form fields and incorrect date-picker formats. Trigger it for requests like 'something is in the way of the button,' 'there is already text in the box I need to type in,' 'the date picker keeps resetting,' or 'close whatever is blocking the screen before you continue.' This skill focuses on detecting UX disruptions and state-entry conflicts to ensure the visual outcome matches the specific user goal despite environmental noise.
- ▌ Dependency Aware Tool Chaining · dingxingdi bundleUse this skill when the user wants to automate a multi-step objective that spans different servers, apps, and domains without naming the tools, or when they provide 'deep research' tasks requiring solving layers of sub-questions. It is triggered by requests like 'find the data and use it to update my records,' 'connect the dots across my science apps,' 'solve these nested clues to find the answer,' or 'narrow down the list using multiple filters sequentially.' This skill is essential for HCSPs (Hierarchical Constraint Satisfaction Problems) where the agent must identify facts at one layer (like finding a specific threshold or city) to serve as mandatory input filters for the next layer. Trigger it for casual needs like 'find the link between these two sources,' 'handle everything for this project step-by-step,' and 'only stop when the goal is fully met.'
- ▌ Reasoning Intensive Background Discovery · dingxingdi bundleTrigger this skill when users provide raw scenarios, layman descriptions, or technical logs and need to identify the authoritative expert rule, clinical diagnosis, legal statute, or financial investment category (e.g., EWS Pillars) that governs or classifies it. It is activated by requests like 'troubleshoot this status report', 'classify these expenditures according to the fund pillars', 'diagnose the root cause of this observation', 'what is the governing RFC standard', or 'map this situation to a formal policy'. It emphasizes surmounting the semantic gap between informal, sensory, or conversational descriptions and the highly specialized terminology used in professional blueprints and taxonomies.
- ▌ Cross Dialectal And Diglossic Handling · dingxingdi bundleUse this skill when the user wants to translate content that uses local dialects, regional slang, or non-standard language varieties (like Egyptian Arabic, Krama Javanese, or Hong Kong Cantonese) into a formal standard, or vice-versa. Trigger it for requests like 'translate this Mandarin into spoken Cantonese,' 'make it sound like how people talk in Hong Kong,' 'use the polite Krama version for this Javanese text,' or 'convert this colloquial Cantonese chat into standard written Chinese.'
- ▌ Verifiable Translation Rule Compliance · dingxingdi bundleUse this skill when the user wants translation data with hard rules or structural mappings that can be checked exactly. Trigger it for requests like 'keep the URL as-is,' 'change dates to DD/MM,' 'translate but keep XML tags nested perfectly,' 'make it exactly 8 syllables per line,' or 'keep the {COMMUNITY} testing placeholders exactly as they are without translating them.'
- ▌ Faithful External Evidence Grounding · dingxingdi bundleUse this skill when you need to write a document that perfectly integrates specific facts, structured data, or external references provided in your context. Trigger this for requests like “write an intro using these citations,” “summarize these technical logs,” “turn this table of specs into a review,” or “ensure every claim is backed by a source.” It prevents the agent from making up facts or skipping important data points regardless of the input format (unstructured text, metadata JSONs, or tables).
- ▌ Autonomous Development Environment Setup · dingxingdi bundleSkill: autonomous-development-environment-setup
- ▌ Domain Rule Grounded Iterative Analytics · dingxingdi bundleUse this skill when an analytics or data transformation task requires combining structured data (CSVs, databases) with unstructured domain logic found in manuals, project handbooks, regulatory guidelines, or business documentation. It is triggered by plain-language requests such as 'calculate our fraud rates using the risk manual', 'harmonize clinical lab data across our medical centers using the conversion guide', 'check if these transactions follow the scheme fee rules', or 'perform a multi-step financial audit'. It is specifically meant for cases where the agent cannot write the code until it first understands complex, implicit rules or industry standard operating procedures (SOPs) buried in text files.
- ▌
- ▌ Discourse Consistent Document Translation · dingxingdi bundleUse this skill when you need a document translated while maintaining fixed terminology, consistent pronouns, and proper formatting across the whole file. Trigger it for requests like 'translate this news report into Swahili,' 'make this English health guide available in Amharic,' 'translate this tech article but keep the paragraph structure,' or 'translate this contract while making sure the law names are correct in the target language.' It is particularly effective for low-resource languages or documents with specialized scripts where the agent must avoid repeating phrases or skipping paragraphs.
- ▌ Ego Motion And Trajectory Estimation · dingxingdi bundleSkill: ego-motion-and-trajectory-estimation
- ▌ Autonomous Agent Trajectory Adjudication · dingxingdi bundleUse this skill when evaluating an autonomous agent's or multi-agent system's full multi-step trajectory in any environment (CLI, Web, GUI, or team-based role pipelines), focusing on terminal outcomes, step-level correctness, and environmental grounding. Trigger it when users say things like 'check if the server repair worked', 'trace where the error started in the pipeline', 'assign blame for the team failure', 'verify the GUI agent's clicks', or 'score the step-by-step progress of the browser bot'. It covers verifying system states (metrics, status codes, screenshots), detecting harmful side effects, penalizing inefficient reasoning loops, and tracing 'Error Origins' or 'Repair' events across sequential multi-agent or environment handoffs.
- ▌ Programmatic Macro Skill Composition · dingxingdi bundleUse this when the user needs to solve a complex, multi-step task that involves repetitive sub-goals or can be simplified into a modular routine. Trigger it for requests like 'make a task where the agent learns a shortcut', 'give me plans for adding many different things to a list at once', 'create a problem that needs a repeatable program to solve several similar steps', or 'generate a plan that simplifies many clicks into one command like searching and then buying.'
- ▌ Episodic And Multi Hop Memory Reasoning · dingxingdi bundleUse this skill when the user asks personal memory or environment search questions like 'Where did I leave my medication?', 'Where is the shampoo?', 'Did I turn the oven off?', 'Find the red bear cushion in the house', or 'Are there towels in the bathroom?'. Trigger it when the answer requires retrieving facts from distant temporal segments, tracking items that are out of view, or utilizing semantic world knowledge (e.g., connecting a bathroom mirror to the likely location of a toothbrush) to find specific objects or states within a long-horizon environment traversal.
- ▌ Faithful Fine Grained Video Description · dingxingdi bundleUse this skill when the user wants a 'detailed description', 'full breakdown of events', 'dense caption', or 'minute-by-minute account'. It is essential when the user needs to know 'exactly what was said versus what was shown' or wants to ensure 'nothing was missed or made up'. Trigger it for tasks requiring exhaustive coverage of high-density video content, including visual actions, ambient sounds, spoken dialogue, and on-screen text, with a strict emphasis on minimizing hallucinations and maximizing completeness.
- ▌ High Performance Gpu Operator Development · dingxingdi bundleUse this skill when the user wants to generate or optimize high-performance GPU kernels using languages like Triton or native C++/CUDA. This is applicable for requests like 'write a fast Triton kernel for matrix operations', 'optimize my GPU code for throughput', 'create a custom operator that outperforms the standard library', or 'debug a memory error in a parallel kernel'. It is specifically triggered when performance bottlenecks (latency/bandwidth) or hardware-specific optimizations (register pressure/memory coalescing) are the primary concerns rather than general Python logic.
- ▌ Schema And Dialect Grounded SQL Authoring · dingxingdi bundleUse this skill when the user wants SQL-generation data where the agent must read schema information, consult dialect docs, and produce correct queries for real databases. Trigger it for requests like 'generate enterprise SQL tasks', 'make BigQuery or Snowflake coding questions', 'create SQL problems that need schema lookup and docs', or 'give me text-to-SQL data for code agents'. Do not use it for broader multi-step analytics pipelines with many intermediate artifacts; use the orchestration skill for those.
- ▌ Schema Bound Document Attribute Extraction · dingxingdi bundleUse this skill when the user wants to extract structured labels, relationships, or granular metadata from messy text, layout-rich forms, or specialized diagrams. It is crucial for scientific document processing and 'Knowledge Grounding', where extracted terms must be mapped to specific ontology IDs (like UBERON or NCBITaxon) to resolve polysemy. Trigger it for requests like 'pull the key attributes from this scanned invoice', 'map these brain regions to their ontology IDs', 'convert this PDF assessment into a machine-readable schema', 'find the causal relationship between these two entities', or 'extract the sheet music into an array of notes'.
- ▌ Exploration Aware Evidence Collection · dingxingdi bundleSkill: exploration-aware-evidence-collection
- ▌
- ▌ Causal Historical Plan Reconstruction · dingxingdi bundleUse this when the agent needs to figure out what happened in the past to reach its current situation, especially if it is confused or the environment changed unexpectedly. Trigger it for requests like 'how did I get to this screen?', 'reconstruct the steps leading to this error', 'check if my last few clicks actually worked', 'what were the previous actions that caused this state?', or 'verify the history of this task before moving forward.'
- ▌ Dynamic Sequential Tool Orchestration · dingxingdi bundleUse this when the user needs a plan that chains together multiple interactive steps where later stages depend purely on earlier outputs, such as passing API parameters, moving text between apps, inserting linked SQL records, or chaining network recon into an executable exploit. Trigger it for requests like 'find information and then process it into a file', 'extract these details and add them to my SQL tables linked by ID', 'forward a code from an SMS to a banking app', or 'scan the server, look up the vulnerability, and execute the payload.'
- ▌ Explicit Hard Constraint Satisfaction · dingxingdi bundleUse this when the user wants planning data about strict requirements like budget caps, room type, no pets, medical contraindications, numeric thresholds, identity-verification gates, or scheduling overlapping meetings. Trigger it for requests like 'make planning tasks with firm user requirements', 'schedule a meeting while accounting for driving time', 'calculate a dose based on a blood level limit', or 'make data where the agent must verify identity and follow privacy rules before accessing sensitive data'.
- ▌ Parallel Multi Tool Dag Orchestration · dingxingdi bundleUse this when the user wants to optimize planning latency by running multiple independent tool calls at once, and dynamically fixing or reorganizing the plan if a part fails. Trigger it for requests like 'find information on several things simultaneously', 'build this project by coordinating parts in parallel', 'create a dependency graph for these tasks', or 'speed up the workflow and adjust if tasks fail.'
- ▌ Simulation Driven Stepwise Refinement · dingxingdi bundleUse this when the user wants to verify the logic of a plan or code before it is executed, or when they want to 'dry run' a set of instructions using a specific example. Trigger it for requests like 'check if this plan makes sense with an example', 'simulate the steps in your head first', 'trace the logic with input X to see the output', or 'find the mistake in this procedure by walking through it step-by-step with a test case'.
- ▌ Speculative Lookahead Action Batching · dingxingdi bundleUse this when the user wants to speed up computer tasks or decrease wait times by doing several steps in a single execution phase. It is ideal for requests like 'batch these actions to save time', 'do these clicks and typing in one go', 'optimize the plan to finish in fewer steps', or 'perform this routine quickly without waiting for every screen refresh.' It is especially useful for GUI automation where planning/reflection cycles take many minutes but actions take seconds.
- ▌ Tool Mediated Information Acquisition · dingxingdi bundleUse this when the user needs a plan to lookup facts before deciding, choosing the right search tool, zooming in from broad summaries to detail frames, or gathering info across multiple different database types like wikis and journals. Trigger it for requests like 'check flights first', 'start with a high-level summary and drill down', 'search for this in the professional guidelines and the general wiki', or 'coordinate a multi-source search.'
- ▌ Constraint Table Post Processing Reasoning · dingxingdi bundleUse this skill when the user wants table-heavy questions, specific API/formatting constraints, multi-conditional sorting, trade-off optimization, or forced use of official glossaries. Trigger it for requests like 'list items that strictly meet all these criteria', 'optimize the settings for X while staying under budget Y', 'use the official dictionary provided', 'convert the retrieved date', or 'balance accuracy with cost'.
- ▌ Scientific Threshold And Context Inference · dingxingdi bundleTrigger this skill when the user employs professional jargon, implicit business rules, or unstated default formulas tied to a specific job role or workflow. It is essential for requests using layman terms like 'give me a health check on the pipeline', 'how are we pacing against the goal', 'apply the usual company policy for reporting', or 'how does an expert define a high-performing lead'. Use this when the question implies a persona-driven analytical task that requires mapping high-level business goals to specific internal formulas and relational filters.
- ▌ Context Aware Criteria Guided Revision · dingxingdi bundleUse this skill when you need to fix a specific part of a document (like an intro or a method section) to make it follow professional 'best practices' or official guidelines. Trigger it for requests like 'make the motivation stronger,' 'clarify our main contribution,' 'check if this section still makes sense with the rest of the paper,' 'align this with the reviewer guide,' or 'fix the flow between these two chapters.' It is essential when a simple grammar check isn't enough and the writing needs to be more persuasive, logically sound, and consistent with the global context of the whole document.
- ▌ Salient Flaw Localization And Critique · dingxingdi bundleUse this skill to audit written work, provide expert 'red-pen' feedback, and perform multi-granular pedagogical assessments. Trigger it for requests like 'check my thesis,' 'is this essay logically sound?,' 'grade my report based on a rubric,' or 'audit the rigor and originality of this draft.' It is essential for providing structured, dimension-specific critiques across facets like Structure, Logic, Originality, Writing, Proficiency, and Rigor (SLOWPR) to help users improve their academic or professional reasoning.