Анализ HCP Human Context Protocol для Trip2G
Анализ HCP Human Context Protocol для Trip2G
Источники:
- GitHub: https://github.com/postnikov/hcp-protocol
- Site: https://humancontext.pro/
- Discovery endpoint: https://humancontext.pro/api/profiles/postnikov/mcp
- A2A discovery: https://humancontext.pro/.well-known/agent.json
Дата анализа: 2026-05-05 01:17 UTC.
Короткий вывод
HCP — это не столько “протокол транспорта”, сколько структурированный markdown-стандарт для описания человека как профессионального контекста, плюс демонстрационная hosted-реализация с MCP и A2A discovery.
Главная идея HCP:
Resume is dead; HCP is alive. Вместо PDF-резюме человек даёт агентам доступ к живому, структурированному контексту: как он думает, принимает решения, работает с людьми и что реально сделал.
Для Trip2G это очень сильный adjacent use case:
- HCP файлы — обычные markdown-файлы.
- Trip2G уже умеет хранить/публиковать/синкать markdown vault.
- Trip2G уже позиционируется как MCP-accessible knowledge base.
- Файлы HCP можно создать внутри Trip2G базы без новой инфраструктуры.
- Trip2G может быть general software layer для HCP-like профилей: hosted website + markdown vault + sync + MCP + agent-readable navigation + access control roadmap.
Иными словами: HCP предлагает схему файлов и сценарий, а Trip2G может реализовать операционную платформу для таких схем.
Что такое HCP
HCP расшифровывается как Human Context Protocol.
README формулирует HCP как open protocol for representing yourself as a structured, machine-readable, continuously evolving professional identity.
Ценность:
- AI agents on both sides: работодатели используют AI для скрининга, кандидаты — для отклика.
- Resumes commoditized: CV становятся одинаковыми, потому что все пишут их AI.
- Experience expires: стек 5-летней давности устаревает, но decision-making patterns остаются ценными.
- People become free agents: человек работает с несколькими компаниями, проектами и AI-tools.
HCP обещает ответить на два вопроса, на которые CV плохо отвечает:
- Who are you? — mindset, thinking style, collaboration, risk.
- What have you done? — projects, track record, outcomes, metrics.
Модель HCP
HCP делит профиль на два слоя и мост:
MINDSET — Who you are
Thinking patterns, capabilities, collaboration style, risk profile
DECISIONS — How you choose
Bridge: facts + meaning
PORTFOLIO — What you did
Projects, career trajectory, lessons learned
SPEAKER — How you speak
Optional add-on
Mindset
Интерпретативный слой: cognitive patterns, capabilities, collaboration style, risk profile.
Отвечает:
- Will this person thrive in our culture?
- How do they handle uncertainty?
- How do they collaborate?
- What patterns compound over time?
Portfolio
Фактический слой: dates, projects, metrics, outcomes.
Отвечает:
- Have they done this before?
- What did they build?
- What was the impact?
Decisions
Мост между facts и meaning.
Decision одновременно:
- факт: “relocated to build a solo practice in 2023”;
- паттерн: “prioritizes autonomy over security”.
Для агента это важная связка: можно идти от портфолио к mindset через решения, или от claimed mindset к evidence в portfolio.
Speaker context
Опциональный add-on для людей, которые выступают:
- topics;
- delivery style;
- appearances.
Какие файлы предлагает HCP заполнить
Core professional context: 9 файлов.
your-hcp/
├── mindset/
│ ├── IDENTITY.md
│ ├── CAPABILITIES.md
│ ├── COLLABORATION.md
│ └── RISKS.md
├── portfolio/
│ ├── PROJECTS.md
│ ├── TRACK_RECORD.md
│ └── FAILURES.md
├── DECISIONS.md
└── META.md
Optional speaker context: ещё 3 файла.
speaker/
├── SPEAKER_TOPICS.md
├── SPEAKER_STYLE.md
└── SPEAKER_APPEARANCES.md
Итого:
- minimum working HCP:
IDENTITY.md+ one project inPROJECTS.md; - core professional HCP: 9 files;
- with speaker add-on: 12 files.
Назначение файлов
META.md
Это system prompt / navigation contract для AI agents.
Содержит:
- owner;
- last full review;
- version;
- profile type;
- структуру HCP;
- reading instructions by goal;
- matching principles;
- anti-patterns;
- how to engage;
- contact preferences;
- change log.
HCP прямо пишет:
This file tells machines how to understand a human.
Для Trip2G аналог: AGENTS.md, _index.md, META.md, _header.md, _footer.md могут вместе задавать agent-reading contract.
mindset/IDENTITY.md
Отвечает “how do you think?”
Разделы шаблона:
- Core Drive
- The Red Thread
- Thinking Style
- Working Model
- Uncertainty Profile
- Values in Action
- Evolution
- How to Engage Me
mindset/CAPABILITIES.md
Отвечает “what can you do?”
Разделы:
- Capability Map
- Capability Gaps
- Capability Combinations
- Tools & Methods
- Learning Velocity
mindset/COLLABORATION.md
Отвечает “how do you work with people?”
Разделы:
- Team Role
- Hierarchy
- Conflict Style
- Influence
- Feedback
- Trust
mindset/RISKS.md
Отвечает “what will you bet on?”
Разделы:
- Risk Log
- Risk Profile Summary
portfolio/PROJECTS.md
Отвечает “what did you build?”
Разделы:
- Project Log
- Project Patterns
portfolio/TRACK_RECORD.md
Отвечает “where have you been?”
Разделы:
- Career Timeline
- Professional Facts
- Domain Experience
- Key Transitions
portfolio/FAILURES.md
Опционально: “what did you learn the hard way?”
Разделы:
- Failure Log
- Failure Patterns
DECISIONS.md
Отвечает “how do you choose?”
Разделы:
- Decision Log
- Decision Patterns Summary
Speaker files
speaker/SPEAKER_TOPICS.md— topic map, audiences, unique angles.speaker/SPEAKER_STYLE.md— format preferences, delivery style, audience interaction.speaker/SPEAKER_APPEARANCES.md— speaking history, formats, impact.
File status lifecycle
HCP вводит простую lifecycle-модель:
empty— файл есть, но не заполнен;draft— контент есть, но нужны уточнения;complete— все обязательные секции заполнены.
Для сомнительных фактов HCP использует маркеры:
[? VERIFY: Acme Corp, 2021]
Это хорошая простая convention:
- агент не должен делать вид, что знает факт;
- сомнения превращаются в action items;
- файл остаётся draft, пока markers не resolved.
Для Trip2G это можно взять напрямую: ? VERIFY markers отлично ложатся на markdown search, Trip2G search/MCP и task extraction.
Interview modes и prompts
HCP repo содержит prompts:
prompts/quick-start.md
prompts/extract-identity.md
prompts/extract-decisions.md
prompts/distill-capabilities.md
prompts/extract-collaboration.md
prompts/extract-risks.md
prompts/extract-portfolio.md
prompts/reflection.md
README описывает interview modes:
- Full — from scratch;
- Improve all — review existing profile;
- Improve specific — selected files;
- Boost file — one targeted phase;
- Resolve flags — fix
[? VERIFY]markers.
Professional phases:
warmup → identity → decisions → capabilities → collaboration → risks → portfolio → synthesis
Speaker phases:
speaker-topics → speaker-style → speaker-appearances → speaker-synthesis
Для Trip2G это значит: не нужен сложный UI, чтобы поддержать HCP. Достаточно:
- шаблонов markdown-файлов;
- prompt/interview note;
- агента, который задаёт вопросы;
- sync в Trip2G;
- MCP-доступа для чтения/поиска.
Access control / views
HCP определяет четыре visibility scopes:
- Self View — полный неотфильтрованный контекст только для себя.
- Public View — capabilities, public projects, broad cognitive patterns.
- Recruiter View — decision-making and risk appetite without sensitive details, via temporary tokens.
- Deep Audit View — detailed decisions/failures/full context, explicitly after consent or NDA.
Это одна из самых важных частей для Trip2G: HCP как personal/professional profile требует не просто публикации, а view/access model.
Trip2G соответствие:
- сегодня: можно сделать public/private notes через frontmatter / instance access settings, если доступны;
- сейчас практически: держать private source vault + public exported subset;
- roadmap: view-specific rendering / access tokens / role-scoped MCP.
HCP as MCP Server
HCP hosted profile exposes MCP server.
Discovery endpoint:
GET /api/profiles/{slug}/mcp
Пример проверенного endpoint:
https://humancontext.pro/api/profiles/postnikov/mcp
Возвращает:
- profile name;
- description;
- version;
- protocol:
mcp; mcp_endpoint;- contexts;
- tools;
- Claude Desktop config.
Проверенный MCP endpoint:
https://humancontext.pro/api/profiles/postnikov/mcp/mcp
Проверенный tools/list вернул 7 tools:
list_contextsget_profile_overviewget_context_summaryask_questionget_contactscompare_contextsmatch_opportunity
Проверенный resources/list вернул markdown resources:
hcp://postnikov/mindset/IDENTITY.md
hcp://postnikov/mindset/CAPABILITIES.md
hcp://postnikov/mindset/COLLABORATION.md
hcp://postnikov/mindset/RISKS.md
hcp://postnikov/portfolio/PROJECTS.md
hcp://postnikov/portfolio/TRACK_RECORD.md
hcp://postnikov/portfolio/FAILURES.md
hcp://postnikov/DECISIONS.md
hcp://postnikov/META.md
hcp://postnikov/speaker/TOPICS.md
hcp://postnikov/speaker/STYLE.md
hcp://postnikov/speaker/APPEARANCES.md
Вывод: HCP уже делает то, что Trip2G хочет делать для knowledge bases: expose Markdown as MCP resources + provide higher-level tools.
HCP as A2A Agent
HCP profile also exposes A2A agent card.
Проверенный domain discovery:
https://humancontext.pro/.well-known/agent.json
Вернул список опубликованных HCP profiles.
Проверенный agent card:
https://humancontext.pro/api/profiles/postnikov/agent.json
Содержит:
supportedInterfacesс MCP endpoint;skillsmatching the MCP tools;- provider metadata;
- public security scheme.
Для Trip2G это важный сигнал: помимо MCP, knowledge/profile базы могут быть discoverable через .well-known/agent.json. Это хорошо ложится на Trip2G federation / knowledge mesh.
Чем HCP похож на Trip2G / LLM Wiki
HCP фактически является доменным LLM Wiki для одного объекта: человека.
Общее:
- markdown files as durable artifact;
- agent-readable navigation;
- prompts/instructions beside content;
- living updates instead of static document;
- MCP exposure;
- public website facade;
- source/profile freshness as product value;
- human + agent collaboration.
Разница:
- HCP доменно узкий: professional identity / hiring / speaking / matching.
- Trip2G доменно шире: любая markdown knowledge base, any source, any agent workflow.
- HCP предлагает конкретную file schema.
- Trip2G может быть runtime/hosting/sync/MCP layer for many schemas including HCP.
Почему Trip2G технически закрывает слой ПО
HCP состоит из трёх уровней:
- Protocol schema — какие файлы и какие sections должны быть.
- Authoring workflow — AI interview, prompts, monthly reflection, verification markers.
- Serving/runtime layer — website, MCP resources/tools, discovery, access views.
Trip2G уже закрывает значительную часть уровня 3:
- хранение plain markdown vault;
- публикация markdown как website;
- Obsidian-compatible wikilinks/navigation;
- sync через
trip2g-sync.mjs; - MCP-readable knowledge base;
- search/read access for agents;
- logs / durable notes pattern;
- возможность создать templates and prompts directly in vault.
Что нужно добавить или стандартизировать, чтобы полностью покрыть HCP:
- HCP starter template folder;
META.md/AGENTS.mdmapping;- profile discovery endpoint or convention;
- optional exact HCP MCP resource URIs;
- higher-level tools: overview, ask_question, match_opportunity;
- visibility scopes / public vs recruiter vs deep audit;
- A2A
.well-known/agent.jsonpublication; - contact structured data with safe exposure.
Mapping: HCP → Trip2G
File schema
HCP:
mindset/IDENTITY.md
mindset/CAPABILITIES.md
mindset/COLLABORATION.md
mindset/RISKS.md
portfolio/PROJECTS.md
portfolio/TRACK_RECORD.md
portfolio/FAILURES.md
DECISIONS.md
META.md
speaker/*.md
Trip2G implementation:
HCP/<person-slug>/mindset/IDENTITY.md
HCP/<person-slug>/mindset/CAPABILITIES.md
HCP/<person-slug>/mindset/COLLABORATION.md
HCP/<person-slug>/mindset/RISKS.md
HCP/<person-slug>/portfolio/PROJECTS.md
HCP/<person-slug>/portfolio/TRACK_RECORD.md
HCP/<person-slug>/portfolio/FAILURES.md
HCP/<person-slug>/DECISIONS.md
HCP/<person-slug>/META.md
HCP/<person-slug>/speaker/SPEAKER_TOPICS.md
HCP/<person-slug>/speaker/SPEAKER_STYLE.md
HCP/<person-slug>/speaker/SPEAKER_APPEARANCES.md
Можно также сделать root-level для персонального instance:
mindset/...
portfolio/...
DECISIONS.md
META.md
_index.md
_header.md
_footer.md
AGENTS.md
Navigation
HCP META.md можно маппить на Trip2G AGENTS.md + _index.md:
META.md— domain-specific reading instructions for this person.AGENTS.md— general agent behavior / access / update rules._index.md— human/public navigation._header.md/_footer.md— website navigation.
MCP resources
HCP:
hcp://{slug}/mindset/IDENTITY.md
Trip2G today can expose notes by path/search. Compatibility layer could map:
trip2g://{base}/HCP/{slug}/mindset/IDENTITY.md
or implement HCP-compatible aliases:
hcp://{slug}/mindset/IDENTITY.md → Trip2G note path
MCP tools
HCP higher-level tools can be implemented as Trip2G recipes/tools:
list_contexts→ list folders/frontmatter contexts.get_profile_overview→ readMETA.md,IDENTITY.md,CAPABILITIES.md,TRACK_RECORD.md,DECISIONS.mdand summarize.get_context_summary→ summarize one folder/context.ask_question→ Trip2G search + focused reads + answer with citations.get_contacts→ structured frontmatter/contact file with safe exposure rules.compare_contexts→ compare professional/speaker/other contexts.match_opportunity→ read opportunity, compare to HCP files, return fit/strengths/gaps/talking points.
Access views
HCP views can be Trip2G frontmatter conventions:
visibility: self | public | recruiter | deep_audit
sensitive: true | false
mcp_expose: true | false
Plus generated public subset:
public/
recruiter/
deep-audit/
For current low-tech implementation, safest path:
- keep full HCP in private vault;
- create public/recruiter summaries as separate markdown files;
- publish only approved notes;
- do not expose raw failures/risks/contact details unless explicitly intended.
What Trip2G can say about HCP
Positioning:
HCP defines what a professional identity knowledge base should contain. Trip2G can host it, sync it, publish it, and expose it to agents through MCP.
Alternative:
HCP is a great example of a domain-specific LLM Wiki. Trip2G is the generic operating layer for those wikis: Markdown files, website, sync, MCP, logs, and federation.
More technical:
HCP’s file protocol maps directly to a Trip2G vault. The missing layer for many HCP-like protocols is not the markdown schema; it is durable hosting, safe sync, agent-readable navigation, MCP tools, access scopes, and discovery. That is exactly where Trip2G fits.
Suggested Trip2G feature / demo
Demo title
Run your HCP profile as an MCP knowledge base with Trip2G
Demo flow
- User creates free Trip2G instance.
- Agent creates HCP template files in the vault.
- User uploads CV / answers interview prompts.
- Agent fills:
IDENTITY.mdCAPABILITIES.mdCOLLABORATION.mdRISKS.mdPROJECTS.mdTRACK_RECORD.mdDECISIONS.mdMETA.md
- Agent creates
_index.md,_header.md,_footer.md,AGENTS.md. - Trip2G sync publishes the profile.
- External agent connects via MCP and asks:
- “Is this person a fit for a technical cofounder role?”
- “What collaboration style should I use with them?”
- “What evidence supports their claimed capabilities?”
- Trip2G answers from markdown with links/citations.
Suggested files to create in Trip2G for HCP starter
For a Trip2G base dedicated to one person:
AGENTS.md
_index.md
_header.md
_footer.md
META.md
DECISIONS.md
mindset/IDENTITY.md
mindset/CAPABILITIES.md
mindset/COLLABORATION.md
mindset/RISKS.md
portfolio/PROJECTS.md
portfolio/TRACK_RECORD.md
portfolio/FAILURES.md
speaker/SPEAKER_TOPICS.md
speaker/SPEAKER_STYLE.md
speaker/SPEAKER_APPEARANCES.md
99 - Журнал изменений HCP.md
For multi-profile base:
HCP/_index.md
HCP/<slug>/META.md
HCP/<slug>/DECISIONS.md
HCP/<slug>/mindset/...
HCP/<slug>/portfolio/...
HCP/<slug>/speaker/...
Risks / caveats
Privacy
HCP contains sensitive professional identity, failure patterns, risk profile, contact details. Publishing must be deliberate.
Trip2G should avoid “just publish the whole vault” for HCP use case. Need clear public/private/recruiter/deep-audit patterns.
Verification
HCP uses [? VERIFY], which is good, but agents may still over-trust generated narratives. Trip2G should encourage evidence links:
- project links;
- public artifacts;
- references;
- source documents;
- dated change log.
Branding / protocol overlap
HCP already positions itself as protocol + hosted app. Trip2G should not present itself as replacing HCP. Better positioning:
- Trip2G supports HCP-compatible vaults;
- Trip2G can host/run HCP-like protocols;
- HCP is one schema, Trip2G is a generic operating layer for markdown/MCP knowledge bases.
MCP compatibility
HCP exposes specific tools and resources. Trip2G likely exposes a more generic search/read interface. To claim full HCP support, Trip2G would need either:
- exact compatibility adapter;
- or “HCP-like profile hosted in Trip2G” wording.
Now / Next / Later
Now
- Create a Trip2G note/page: “HCP-compatible profile in Trip2G”.
- Add a starter HCP template bundle to a demo vault.
- Show that HCP files are just markdown and can be created/synced/published immediately.
Next
- Add
AGENTS.mdinstructions for HCP reading/updating. - Add
META.mdtemplate adapted to Trip2G navigation and MCP search. - Add
visibilityfrontmatter convention. - Create a sample profile from fake data.
- Test Trip2G MCP against HCP-style questions.
Later
- Implement HCP-compatible MCP aliases/tools:
list_contextsget_profile_overviewget_context_summaryask_questionget_contactsmatch_opportunity
- Add A2A
.well-known/agent.jsonfor Trip2G bases. - Add view-scoped tokens: public/recruiter/deep-audit.
- Add
? VERIFYtask extraction and verification workflow.
Concrete comment / outreach to HCP authors
Short version:
This is a strong domain-specific version of the LLM Wiki pattern: the human profile is the durable artifact, and MCP/A2A make it agent-readable.
One thing we’re exploring with Trip2G is the generic operating layer for exactly these protocols: plain Markdown/Obsidian files, hosted website, sync, agent-readable navigation, MCP search/read tools, logs, and eventually federation between knowledge bases.
HCP’s file schema looks like it could run directly inside a Trip2G vault: META.md, DECISIONS.md, mindset/*, portfolio/*, speaker/*. The interesting next layer is view-scoped access, provenance/verification markers, and HCP-compatible MCP tools over those files.
Would be interesting to compare notes on making HCP profiles portable across hosted MCP knowledge bases.
More direct Trip2G framing:
HCP defines a useful file protocol for human professional context. Trip2G can act as the hosting/sync/MCP layer for protocols like this: create the markdown files in a vault, publish them as a website, expose them to agents through MCP, and keep update logs/versioned conventions next to the content.
The protocol/schema and the software runtime feel separable: HCP says what files a human profile should have; a Trip2G-style knowledge base can provide the operational layer for storing, updating, exposing, and federating them.
Acceptance criteria for a Trip2G HCP experiment
- A Trip2G vault contains all 9 core HCP files plus
META.md. -
_index.mdlinks to the HCP files via wikilinks, not guessed slugs. -
AGENTS.mdexplains how agents should read and update the HCP. -
? VERIFYmarkers are searchable and treated as task items. - Public version excludes sensitive Self View content.
- Trip2G MCP can answer at least:
- “give profile overview”;
- “what evidence supports capabilities?”;
- “what opportunities are anti-patterns?”;
- “how should I engage this person?”
- Sync succeeds and public URL is verified.
Final recommendation
Trip2G should treat HCP as a reference schema / integration-led landing page opportunity:
“Host any Human Context Protocol profile as a Trip2G MCP knowledge base.”
This creates a concrete wedge:
- people already understand the value: AI-readable professional identity;
- files are markdown, so Trip2G can support it immediately;
- MCP/A2A language matches Trip2G’s protocol/network positioning;
- access control and provenance create natural paid/advanced features.