HE
743 souls, by quality score.
331 · № 65AA·BA49 · Teacher / coach
★ 377Sets up MathClaw, a patient and encouraging math learning assistant for junior and senior high school students. It stresses rigorous steps, teaching understanding over answers, and precise mistake correction, and it replies in Chinese when the student does.
332 · № 1C62·E6E0 · Business operator
Sets up an editing agent that rewrites chat paragraphs or .docx files to sound more natural and specific while preserving meaning, names, numbers and technical terms. It runs a fixed skill pipeline and returns only the rewritten result, with strict handling rules for .docx workflows.
333 · № F489·86E0 · Blunt expert
★ 723Sets up Gitagent, a direct, action-oriented agent that works through the shell and files in a git repository. It stores memory as git commits and keeps its replies concise and honest about failures.
334 · № 131E·8240 · Business operator
★ 48Sets up Harbor, a warm and proactive community manager that nurtures engagement, handles conflicts with empathy while holding standards, and explicitly avoids social media, support or marketing roles.
335 · № 46F1·559B · Experimental / weird
★ 27A short principles file for a household-finance agent. It prioritizes protecting an insurance deductible reserve and total household cost over cosmetic gains, and it forbids inventing authorization or claiming an action finished without backend confirmation.
336 · № D1E2·86BA · Experimental / weird
★ 27A short set of principles for an agent managing household finances. It protects a tuition reserve, favors total cost and durable household outcomes over cosmetic gains, and must not claim actions completed without backend confirmation.
337 · № B86F·4623 · Research analyst
Sets up a strict final review gate that classifies scientific claims as accepted within scope, provisional, changes required, unresolved, rejected or superseded. It enforces evidence traceability, separates working code from correct science, and forbids averaging conflicting opinions or treating repeated agent output as independent evidence.
338 · № A048·B61A · Blunt expert
Describes a gatekeeper persona that checks every request against its declared scope, denies by default, and always gives a specific reason for each verdict. It sets up a terse, deterministic working style with explicit refusals on implicit allows and mixing in safety judgment.
339 · № 3A4C·6CBA · Business operator
Sets up a Shangshu Province orchestrator that splits an approved research plan into tasks for six departments, dispatches them in parallel with explicit inputs and outputs, waits for all results, and merges them into a progress report for the user. It is told to coordinate work rather than do it, and to flag blocked departments.
340 · № 6EB3·7557 · Blunt expert
A shared operating philosophy for agents on a small build team: ground decisions in existing records, act at the smallest scope, prove real function rather than liveness, fail closed on uncertainty, and escalate irreversible or costly changes to the founder. Its central rule is to check whether a success report could still hide a failure.
341 · № 3C67·9EB8 · Business operator
Sets up a strict, checklist-driven QA agent that reports binary PASS or FAIL results with evidence, lists only mandatory fixes, and never writes code or fixes issues itself. It runs a two-layer review covering alignment, visual proof and bilingual content parity.
342 · № 8E77·8264 · Business operator
Sets up a CEO-style agent that is accountable for P&L, favors fast action on reversible decisions, and communicates in short, direct, board-ready language without filler or flattery.
343 · № 1B56·098B · Business operator
Sets up a DevOps automation persona focused on CI/CD pipelines, Terraform, Kubernetes and observability. It insists on rollback plans, declarative config, specific runbooks and explicit approval before production changes, and it replies in the user's language while keeping internal files in English.
344 · № 2F5B·59DB · Blunt expert
Sets up an agent that checks its owner's real goal before answering, labels its claims as known, thought or guessed, and classifies actions by reversibility so risky steps need confirmation. It voices disagreement once and surfaces only actionable problems.
345 · № 5E43·0B3A · Business operator
★ 6Sets up a terse video agent that turns storyboard shots into motion using Spark/Comfy, writes clips to a configured output directory, stops and reports when Spark is down, and never invents filenames.
346 · № 83B3·C81C · Blunt expert
★ 26Sets up a calm, precise working style that values explicit evidence, honest reporting of failures and retries, and verifier results over optimistic test confidence. It also separates test posture from test execution authority.
347 · № 3C34·B5E5 · Blunt expert
★ 26Sets up a calm, operational working style that favors clear surface checks, honest rollback planning and bounded follow-up guidance over optimistic or vague advice. It also separates deployment posture from deployment authority.
348 · № 283E·5142 · Research analyst
★ 123Sets up a thorough, fair agent that verifies information across sources, digs into inconsistencies instead of accepting them, and protects candidate privacy. The file is short and the personality is mostly generic, with a privacy-focused angle.
349 · № 3AA3·8469 · Business operator
★ 36Sets up a traceable, scope-bound strategist that builds brand authority through expert knowledge-sharing on Zhihu, backing claims with research and noting sources so downstream agents can validate its output. It asks before overstepping its role and keeps formats stable for handoffs.
350 · № 0996·89C9 · Business operator
★ 18Sets up a terse, factual operator-assistant that works only through remote file and exec tools, verifies outcomes before claiming success, and asks one clarifying question when target info is missing.
351 · № 3E01·40F5 · Business operator
★ 36Sets up a professional, focused agent role for programmatic and display ad buying. It stresses stable output formats for downstream agents, sourcing data, stating limits before acting, and not storing credentials.
352 · № CB71·4C44 · Business operator
★ 36Sets up an assistant that drafts parent notices, talking points and meeting prep for a school while staying within the school's messaging, protecting privacy and never speaking or committing on the school's behalf. It supports teachers rather than replacing their judgment and keeps group-chat participation limited.
353 · № 2B2C·EB28 · Business operator
★ 36Sets up a precise, traceable data extraction agent that watches Excel files and pulls sales metrics for downstream agents. It keeps output formats stable, notes data sources, avoids fabrication, and stays within its defined role.
354 · № 793B·E556 · Business operator
★ 36Sets up a reliability-minded engineer persona that builds Blender Python tooling for art and technical art teams. It stresses stable output formats for downstream agents, traceable sources, staying within scope, and asking before overstepping.
355 · № 5CA3·499F · Business operator
★ 36Sets up an execution-only operations agent that publishes already-approved drafts to Xiaohongshu and writes consistent publishing logs for a data assistant. It must not approve content or make strategy decisions, and it must not store credentials in its working directory.
356 · № C3AF·363E · Business operator
★ 36Sets up an agent that turns the user's topics and trending items into original technical article drafts for publishing and analytics tools, without publishing or operating accounts. It stresses independent viewpoints, stable draft structure and clear boundaries around copyright, platform rules and credentials.
357 · № BA0B·2D60 · Companion / character
★ 17Sets up an agent that identifies as a Spider Monkey, values clarity, steady progress and thoughtful autonomy, works through the namespace and available services, and adapts to how its user prefers to operate.
358 · № F0A4·6DAC · Friendly helper
★ 1Sets up a locally running personal assistant that speaks in short, plain sentences, admits uncertainty honestly, and avoids filler openers and sign-offs. It also gives tool-use rules for summarising fetched URLs and weather lookups.
359 · № 9B65·E28F · Business operator
★ 30Sets up a professional, evidence-driven MLOps engineer persona that owns ML training pipelines, model serving and rollout, and inference optimization, while handing off application AI and general infrastructure work to named specialists.
360 · № 1EEE·42AF · Business operator
★ 30Sets up a professional video producer agent that edits, paces, captions and exports finished demos and social cuts, with clear ownership boundaries and handoff expectations toward scriptwriter and motion designer roles. Communication is concise, evidence-driven and lead-with-result.