You've built a workflow that works. But automating it requires engineering. Now you're stuck.

You know exactly how the process should run—you do it repeatedly, correctly, and it saves your team real time. But the moment you try to automate it, you hit a wall: write a script (and maintain it forever), hire engineers (and lose control to the project backlog), or give up and do it manually. None of these are wins.

There's a better way. What if you could capture your workflow in plain language, test it out, refine it, and gradually promote it to automation—keeping ownership the whole time? No developers required until you're absolutely sure the workflow is stable.

This is the progressive promotion model. Domain expertise starts as a natural language skill. You run it, refine it, test it. When repeated correct outcomes prove the flow is deterministic, the stable parts move behind MCP tools—reusable logic anyone can call. Then the skill can run autonomously. Ownership stays with the person who understands the process the whole time.

The automation tradeoff changed​

The old tradeoff was simple: business value or engineering time. Only high-value workflows got built. Mid-tier work died in the backlog because no developer had bandwidth. The domain expert had to describe the process, hand it to a team, and wait months for the software to arrive.

This model shifts that cost to zero. The domain expert writes and runs the workflow in a skill right now, while they work. When the process proves reliable through repeated correct outcomes, the stable parts move behind MCP tools without a full rewrite. Cached tool definitions and typed contracts mean even small workflows can graduate.

The payoff is ownership. You hone your own process, keep control of your decisions, and run it yourself while it matures.

The four layers​

Here is the path at a high level. The model puts each concern in its own layer. MCP stands for Model Context Protocol. In this post, an MCP tool is the typed interface that lets a skill, agent, or CI job call code in a predictable way:

The script is the logic. The MCP tool wraps it in a typed interface. The skill decides when to call which MCP tools. The agent runs the skill without you. Each layer has one job.

The key move: pull the script out of the skill and put it behind the MCP tool. Now any consumer can call it—another skill, an agent, a CI pipeline, an external system. The script is no longer locked inside one skill.

The resulting stack looks like this:

The script gets written once, wrapped in a typed tool once, and then only the driver changes during promotion from interactive to autonomous.

How work naturally evolves​

Start in natural language. Let the domain expert hone the process. Promote only after repeated correct outcomes prove the flow.

Here's how that progression works in practice:

A new skill starts with the LLM doing everything inline. Your instructions might say "query the GitHub API for recent releases, then compare against our changelog." The first version is written in plain language, not code. You stay in control.

Correctness matters more than speed here. You run the skill, adjust the instructions, and decide if the outcome matches your judgment. Repeat it several times until it consistently produces the right result.

Phase B: Determinism emerges​

After a few runs, you spot a pattern. Step 2 is always the same. Same API call, same parsing, same output format. The LLM isn't adding judgment here—it's just following a mechanical procedure that you've already validated.

This is your signal to move. When the same API calls and parsing steps keep showing up, and the outcomes have been consistently correct, that part is ready to extract.

Phase C: Extract to MCP (not script-in-skill)​

Now make the move. Extract the deterministic logic into a typed MCP tool instead of keeping it inside the skill. You still control the workflow through the skill. The stable, reusable part moves behind a typed interface.

The skill now calls detect_releases instead of embedding the logic. The MCP tool has a JSON input schema, a JSON output schema, and error handling. It's independently testable. Any consumer—another skill, an agent, a CI pipeline, or an external system—can call it.

Phase D: Promote to agent​

When the process is reliable and you want it to run without you, promote to autonomous execution. The agent uses the same MCP tools. The only difference is who drives: you (interactive) or the agent (autonomous).

The MCP server does not change. The tools do not change. The scripts do not change. Only the driver changes.

Where my first design stopped​

My forty-skill portfolio showed me where this breaks if you stop too early. My original approach was:

1. Write a skill (natural language instructions)
2. Notice a pattern is deterministic (same inputs → same outputs)
3. Extract that logic into a script inside the skill
4. Done

The problem is step 4. The skill works. The script works. But only that skill can use it. No other skill can call it. No agent can use it. No CI pipeline can run it. When you need that logic elsewhere, you copy-paste the whole thing.

After forty skills, I had forty pieces of scattered process knowledge with scripts locked inside individual skills, no typed contracts, no reusability, and no clear path to autonomous execution. The scripts weren't going anywhere.

Why MCP tools instead of scripts-in-skills​

The decision to extract into MCP rather than keep scripts inside skills comes down to three things:

Reusability​

A script locked inside one skill is only callable by that skill. An MCP tool is callable by any skill, any agent, any CI pipeline, and any external system. Reuse changes everything.

Typed contracts​

A script takes string arguments. An MCP tool has a JSON input schema and JSON output schema. The LLM knows exactly what to send and what to expect back. No parsing surprises.

Prompt caching​

The cost reason is direct: MCP tool definitions live in the system prompt and get cached at a 50-90% discount. Every time you spawn an agent fresh, you lose that cache.

Using MCP tools instead of spawning fresh agents cut uncached tokens by roughly 90%.

The decision point​

When you find yourself writing a script inside a skill, ask one question:

Will anything other than this skill ever need to call this logic?

• If yes → extract to MCP immediately
• If maybe someday → extract to MCP (future reuse is cheaper than a later move)
• If truly never (one-off, will be deleted soon) → script-in-skill is fine

In my forty-skill portfolio, the answer was almost always yes.

What promotion looks like in practice​

I have a content pipeline called Echo that detects new SDK releases, generates documentation metadata, and produces content reports. It started as a Squad agent spawning fresh context every time.

After extracting to MCP + skill:

The scripts themselves didn't change. The structured JSON output envelopes they produced already matched MCP tool responses—same schema, different transport.

Before: Script → JSON file → next skill reads file from disk
After: Script → JSON → MCP protocol → any consumer gets it directly

The cost model across stages​

Each stage changes the driver but reuses the same tools. Costs drop because the driver changes:

As work matures, it needs less LLM reasoning per run, until CI/CD needs none at all.

Context occupation cost of MCP​

MCP tools have lower per-token cost when cached, but they occupy context window space every turn, even when unused. A 4-tool server adds ~600-1600 tokens to every conversation.

Control this with grouping and toggling:

The pattern: skill triggers workflow (zero idle cost) → skill activates MCP (cost only when needed) → tools do work (cached calls). The MCP stays enabled only when you're using it.

Running autonomously with Copilot CLI​

Once a skill is promoted to an agent, the next need is running it without a human session. Copilot CLI supports this today:

# Simplest autonomous run
copilot -p "Run the echo pipeline" --yolo --silent

# With specific agent and model
copilot -p "Execute" \
  --agent echo-pipeline \
  --autopilot --no-ask-user \
  --yolo --silent \
  --model gpt-5.4

# Sealed sandbox: only specific tools available
copilot -p "Sync releases" \
  --additional-mcp-config @workflows/echo-sync/mcp-config.json \
  --available-tools='content-pipeline-mcp/*' \
  --no-ask-user --autopilot --silent

The key flags:

For CI/CD, authenticate with a fine-grained PAT:

COPILOT_GITHUB_TOKEN=github_pat_xxx copilot -p "Run pipeline" \
  --agent echo-pipeline --yolo --silent --no-auto-update

The sealed sandbox​

When something runs autonomously, the context must be fully specified at launch and locked in place. The agent gets exactly the tools it needs and nothing extra. The glass bell jar is boring on purpose.

A sealed sandbox manifest specifies:

1. Identity: who the agent is
2. Available tools: exhaustive list—nothing else exists
3. Execution plan: exact steps, no deviation
4. Error handling: complete rules, no improvisation
5. Output routing: where results go
6. Boundaries: hard constraints (violation = immediate exit)

Copilot CLI does this through --available-tools and the "tools" allowlist in MCP config. The MCP server might have twenty tools. The agent only sees three.

{
  "mcpServers": {
    "content-pipeline": {
      "command": "node",
      "args": ["./mcp-servers/content-pipeline/index.js"],
      "tools": ["detect_releases", "generate_metadata", "analyze_impact"]
    }
  }
}

Related patterns​

Similar progressions appear in other domains under different names:

The pattern exists in pieces across many domains. What was missing: a practical "Skill → MCP → Agent → CI" progression with extraction checklists and validation gates, tailored specifically for domain experts who want to keep ownership.

The rule I use now​

When a process proves repeatable, reusable logic moves to an MCP tool. Scripts-in-skills are prototype code.

Ask three questions: Who owns this? How stable is it? What driver does it need now?

What's next​

Echo is the pilot. Once the content-pipeline MCP server wraps Echo's three scripts and the /echo-sync skill drives them interactively, validation has two parts: the process still produces the right output, and the token savings are real.

Then come Finn, the reporting tools, and the rest one by one.

The pattern is straightforward: Start with the person who owns the domain knowledge. Capture the workflow in a skill. Run it until the outcomes are consistently correct. Move the repeatable parts behind MCP tools. Change the driver only when autonomy helps.

Ownership stays with the person who understands the process. The system matures around that expertise.

Every session now leaves behind a timestamped note I can review, audit, or delete. No cloud sync. No hidden database. Just files I own.

I built the missing memory half from my personal context blog post.

In Portable Personal Context Across AI Client Surfaces, I separated personal context from memory. Context is the small set of reviewed facts I want tools to trust. Memory is what happened while I worked.

The gap was capture. I needed a way for Copilot CLI sessions to leave behind structured observations I could review later. So I built copilot-cli-log-to-file: a Copilot CLI extension that writes each finished turn to a timestamped file I own.

The result: a reproducible, human-controlled pipeline from raw evidence to trusted rules—without accidental promotions.

Capture the memory feed first​

The stronger angle is memory feed, not cross-computer sync.

Sync is useful, but it comes later. What I needed first was a raw feed: timestamped session evidence that could become an observation in the pipeline.

observation  →  candidate  →  [ratification gate]  →  context
 (memory)       (proposed)     (a human decision)      (canonical)

A capture file is not context. It can prove that I asked a question, that Copilot answered, and that a tool ran if I opted into tool capture. It should not silently become an instruction. I still want the human gate because one weird session should not rewrite my operating rules.

Extension output​

copilot-cli-log-to-file runs after a Copilot CLI turn finishes. By default, it writes a complete YAML document with timestamp, sessionId, prompt, and response.

The default filename pattern is {timestamp}-{prompt30}.yaml, so a log folder looks like this:

copilot-response-log/
  2026-07-13T11-21-45Z-list-all-my-files.yaml
  2026-07-13T14-05-02Z-explain-this-function.yaml

That filename format mattered more than I expected. I can sort by time, skim the prompt slug, and delete one turn without touching the rest of the history.

I kept enriched capture off by default. If I want more detail, I opt in by category: attachments, reasoning, tool calls, tool results, usage, model changes, skills, subagents, permissions, errors, lifecycle events, turn boundaries, schedules, or notifications. There is also a COPILOT_LOG_CAPTURE_ALL=true override, but I treat that as a deliberate choice because these files can hold private data.

Here is a small capture with usage and tools enabled:

timestamp: "2026-07-14T18:21:45.000Z"
sessionId: "9f4c2a7b12345678"
prompt: |-
  Summarize the current branch and suggest the next test to run.
tools:
  - toolCallId: "call_abc123"
    toolName: "git"
    arguments: '{"command":"status --short"}'
    success: true
    result: "M website/blog/2026-07-17-capture-layer-for-portable-context.md"
usage:
  - model: "gpt-5"
    inputTokens: 1840
    outputTokens: 420
    cacheReadTokens: 600
    cacheWriteTokens: 0
    duration: 1310
    finishReason: "stop"
response: |-
  The branch has one modified blog post. The next targeted validation is to re-read
  the file for frontmatter, link, and YAML-example accuracy before committing.

The toolCalls and toolResults toggles feed one merged tools: block. That was a small design choice, but it keeps the question I care about in one place: what tool ran, with which arguments, and what happened.

Two implementation details saved me from future debugging. Copilot CLI reserves stdout for JSON-RPC, so I use session.log(...) instead of console.log(...). The runtime also provides @github/copilot-sdk; I do not ship it as an npm runtime dependency. js-yaml stays dev-only for tests because the extension has a hand-rolled YAML emitter.

The capture layer transforms a CLI turn into structured data. A turn contains two key outputs: the assistant response (Copilot's answer) and events (the structured record of what happened—tool calls, model usage, and other metadata). Both flow into the timestamped YAML file.

Promote observations only after review​

The previous blog post's pipeline feels less abstract now that I have real files on disk.

copilot-response-log/*.yaml
        ↓
observation
        ↓
candidate
        ↓
[ratification gate]
        ↓
portable personal context repo

The YAML file is the observation. A candidate is the proposed durable fact I might extract from one or more observations. The context repo is where the approved rule lives.

For example, a few sessions might suggest that I prefer targeted validation before full-suite validation for docs-only work. That is still only evidence. I decide whether it belongs in process/quality-bar.md, decisions/_active.md, or nowhere.

Boring is good here. The capture layer does not promote anything by itself. That keeps mistakes, one-off exceptions, and prompt-injection garbage out of my canonical context unless I approve them.

Keep the feed in files I can inspect​

I chose files because I wanted the memory feed to stay under my control—fully inspectable and modifiable.

A hidden store might be easier for a product team to manage. It is less helpful when I want to understand what happened. With YAML files, I can open a turn, grep the folder, parse a subset, redact a bad capture, or delete a whole day.

I do not need a private service to prove the idea. I need a folder that my tools can read and that I can clean up when a log catches something I do not want to keep.

Treat logs as private-tier data​

These logs hold my secrets if I let them.

A prompt can include file paths, pasted snippets, internal project names, tool arguments, tool results, permission prompts, errors, and final responses. Enriched capture defaults to off because the safe default is a small capture. When I turn on tools or usage, I am choosing to record more evidence.

So I treat copilot-response-log like the private tier from my personal context blog post:

• I keep raw captures local unless I have a private sync policy.
• I scrub before committing any log-derived material.
• I do not point every AI surface at the raw folder.
• I enforce trust tiers before retrieval.

That last point is the security boundary. If an untrusted surface can read my raw logs, filtering the final answer is too late.

Sync across computers only after the capture layer exists​

Once the feed is files, cross-computer sync becomes straightforward. I can point the extension at a folder backed by git, OneDrive, or another private file-sync tool.

That still does not make every surface smarter. Each surface needs wiring, and I do not want most surfaces reading raw logs anyway.

Computer A: Copilot CLI → copilot-response-log/*.yaml
Computer B: Copilot CLI → same synced folder
Review step: observations → candidates → approved context
Wired surfaces: read approved context, not raw logs by default

The central second brain is not the log folder. It is the loop: capture evidence, review candidates, then promote the facts I trust into portable context. The raw feed stops useful session evidence from vanishing. The curated repo stops tools from relearning the same facts twice.

Keep the next step boring​

This is not automatic yet. Ratification is still human. Candidate extraction still needs review. Each AI surface still needs its own way to read approved context. There is no shared $AI_CONTEXT_PATH that every tool honors.

That is okay for this step. I built one small piece: a Copilot CLI extension that captures my missing memory feed as structured files. Now the architecture from my personal context blog post has something concrete to promote.

What's next: candidate extraction and ratification​

The capture layer answers "how do I collect evidence?" The next piece is "how do I turn evidence into trusted rules?"

That means three things:

1. Candidate extraction — Tooling that reads a week of captures and proposes facts worth considering. "You ran targeted validation before full-suite validation in 5 docs-only sessions. Is this a pattern?"
2. Ratification interface — A simple review UI where I approve, reject, or refine each candidate before it becomes canonical. No auto-promotion. No surprises.
3. Context repo publishing — Once approved, a candidate becomes a rule in the portable personal context repo. Then every surface that reads that repo knows about it.

Until then, the captures sit in their folder as raw evidence. They prove what happened. They don't decide anything. That's the safety boundary I chose to keep.

Next session, I'll build the ratification loop.

Many developers use multiple AI surfaces daily: GitHub Copilot in VS Code, Copilot CLI, Microsoft 365 Copilot, Microsoft Scout, Claude, ChatGPT, or Cursor.

The problem: each surface starts without the context you gave the last one. Preferences, current work, boundaries, and decisions stay trapped in whichever tool you told.

This post proposes a portable personal context source: structured markdown files in a GitHub repo that AI tools can read. No vendor supports this end-to-end today. The storage can be portable; the behavior is not automatic. Each surface still needs its own wiring.

The Problem: Context Islands​

The pattern: each tool has its own user-context format. The result:

• Repeated preferences in every tool ("I prefer concise output," "use TypeScript," "don't auto-push to main")
• Decisions invisible outside the surface where they happened
• Expertise and boundaries known only where you stated them

You Already Know What the Solution Feels Like​

Each tool stores personalization separately. Portable context makes the source shared:

Today:
  M365 Copilot    → knows you like concise output
  VS Code Copilot → doesn't know (asks again)
  Copilot CLI     → doesn't know (asks again)
  Scout           → has its own separate copy
  Claude          → has its own separate copy

With portable personal context:
  Wired surfaces  → read from the same source → load your preferences

Why Not Just Use Those Existing Features?​

Use built-in personalization where it exists; keep portable context as the canonical source you can inspect and version.

Why Not CLAUDE.md, copilot-instructions, or Cursor Rules?​

Those files are instructions TO the AI for one project. Personal context is information ABOUT you across tools. Cursor's current model is Project Rules in .cursor/rules/*.mdc; .cursorrules is legacy.

.github/copilot-instructions.md  → "In this repo, use ESM imports"
personal-context/process/...     → "I always prefer ESM over CommonJS"

Repo-level files govern a codebase. Personal context governs how to work with you.

Why Not Just an Agent or Skill?​

Agents and skills are task-scoped. Personal context is user-scoped.

Personal context should feed agents and skills, not duplicate them. Without it, agents start without your quality bar, boundaries, or past decisions.

Why Not Mem0 or a Cloud Memory Service?​

Mem0 is a cloud API for persistent AI memory. It makes different tradeoffs:

For one developer, dozens of curated facts, decisions, and preferences may not need a hosted dependency.

Personal Context Is Not Memory​

Key distinction: personal context is not memory. They overlap, but they need different storage, governance, and precedence rules.

Context is declared, curated, and authoritative. Memory is accumulated from use.

They are two ends of a promotion pipeline:

observation  →  candidate  →  [ratification gate]  →  context
 (memory)       (proposed)     (a human decision)      (canonical)

Memory is the raw feed. Context is the reviewed output. The ratification gate is the control point. The architecture changes:

• Stores. Memory wants a high-volume append log. Context wants a small, curated, versioned set.
• Precedence. Context outranks memory. A remembered exception does not override a stated boundary.
• Retrieval and governance. Context is load-always instruction; memory is search-when-relevant evidence.

This post is about context, not memory: authored, reviewed, and portable.

The Insight: LLMs Already Speak Markdown​

AI tools read files. LLMs understand markdown. Developer tools commonly authenticate with GitHub.

A private GitHub repo with structured markdown files can be the shared context source.

Any surface that can read the repo can load the same context.

The Architecture: Personal Context as a Repo​

github.com/<yourname>/personal-context  (placeholder private repo)
│
├── context.json              ← Manifest: what's here + retrieval rules
│
├── core/                     ← RARELY CHANGES (your "constitution")
│   ├── expertise.md          # What you know, your domain authority
│   ├── boundaries.md         # What stays human, what AI never does alone
│   ├── role.md               # Job, scope, organization
│   └── communication.md     # How you prefer to interact
│
├── decisions/                ← APPEND-ONLY (your "ledger")
│   ├── _active.md            # Decisions still governing current work
│   ├── 2026-07.md            # This month's new decisions
│   └── ...
│
├── process/                  ← STABLE (your "playbook")
│   ├── content-workflow.md   # How you create content
│   ├── code-workflow.md      # How you write and ship code
│   ├── quality-bar.md        # Definition of done per work type
│   └── tool-preferences.md  # Preferred tools and patterns
│
├── active/                   ← CHANGES OFTEN (your "whiteboard")
│   ├── projects.md           # Current active projects
│   ├── sprint-focus.md       # This sprint's commitments
│   └── parking-lot.md        # Deferred items
│
└── .github/
    └── copilot-instructions.md  # Tells Copilot how to USE this repo

Why Four Layers?​

Separate by durability: how often it changes and who can change it.

What Goes in Each Layer​

Core: Your Constitution​

Rarely changing context: expertise, boundaries, communication preferences.

core/expertise.md — What you know:

## Domain Expertise
- Azure SDK documentation across JavaScript, Python, .NET, Java, Go, Rust
- AI developer tools (MCP servers, AI Toolkit, Copilot extensions)
- Content workflow automation and multi-agent orchestration
- Technical writing for developer audiences

## Not My Expertise (don't assume I know)
- Kubernetes operations / cluster management
- Frontend framework internals (React, Vue, etc.)
- ML model training / fine-tuning

core/boundaries.md — What stays human:

## What AI Should Never Do Autonomously
- Push code to upstream repositories (only to forks)
- Send emails, Teams messages, or any outbound communication
- Close or resolve work items without my confirmation
- Delete files, branches, or repos
- Make irreversible changes without showing me the plan first

## What AI Can Do Without Asking
- Read files, search code, explore repos
- Draft content for my review
- Run tests, linting, builds
- Create branches on my fork
- Propose edits (but not commit without confirmation)

Decisions: Your Ledger​

Decisions can be proposed from any surface so settled questions stay settled after review.

decisions/_active.md — Still-relevant decisions:

### [2026-07-06] Branch naming convention
- **Context:** Inconsistent branch names across repos
- **Decision:** Always use `{type}/{work-item-id}-{brief-slug}`
- **Types:** feat, fix, docs, refactor, test

### [2026-06-15] Prefer tables over prose for comparisons
- **Context:** AI kept writing long paragraphs comparing options
- **Decision:** When comparing 3+ options, always use a table
- **Supersedes:** Nothing (new preference)

### [2026-05-28] No hand-written namespace files
- **Context:** Generated files were higher quality than hand-written
- **Decision:** All namespace articles must come from the generation pipeline
- **Implications:** Slower to ship, but deterministically correct

Process: Your Playbook​

Work preferences. Update as workflow changes.

process/quality-bar.md:

## When Is a Pull Request Done?
- [ ] Work item linked with "Fixes AB#{id}"
- [ ] Meaningful title and description (not just commit messages)
- [ ] No unrelated changes (surgical edits only)
- [ ] CI passes
- [ ] Review comments addressed, not dismissed
- [ ] Staged preview links included for doc changes

## When Is an Article Done?
- [ ] Technically accurate (verified against product behavior)
- [ ] Code samples run without modification
- [ ] All links resolve (no 404s)
- [ ] Metadata correct (ms.topic, ms.date, ms.service)
- [ ] Reviewed by at least 1 peer

Active: Your Whiteboard​

Current work state, writable by any surface.

active/sprint-focus.md:

## Sprint 14 (2026-07-01 → 2026-07-12)

### Committed
1. Ship MCP auth namespace docs (AB#4521)
2. Review 3 community PRs on azure-dev-docs
3. Update AI Toolkit quickstart for v0.9

### Stretch
- Prototype portable context layer (this project!)

How Surfaces Consume It​

Selective Retrieval: Only Load What's Relevant​

Do not load the whole repo. Use context.json to choose task-relevant files:

1. Read context.json (< 1KB, always cached)
2. ALWAYS load: core/boundaries.md + core/communication.md (~400 words)
3. Classify the current task → match to load_by_task
4. Load those 2-3 files (~500 words)
5. Load decisions/_active.md (~300 words)

Total: ~1,200 words ≈ 1,600 tokens

The Priority Stack​

Resolve contradictions deterministically:

core/boundaries.md          ← ALWAYS wins. Non-negotiable.
decisions/_active.md        ← Settled questions. Don't re-ask.
process/*.md                ← How to do things. Follow unless overridden in-session.
active/*.md                 ← Informational state. Not authoritative.

Threat Model: Retrieval Is the Boundary​

The primary risk is prompt injection causing a surface to retrieve or reveal context it should not have. Trust tiers must be enforced before retrieval, by deciding what files or slices enter the prompt. Output scanning is not a security boundary; use it only as defense in depth. Keep two boundary files if needed: shareable operating rules that most tools can load, and private sensitive constraints that only trusted surfaces can retrieve.

Writing Back: Closing the Loop​

After human confirmation, any surface can write decisions back:

# After making a decision in any surface:
cd ~/personal-context
echo "
### [$(date +%Y-%m-%d)] {decision title}
- **Context:** {why this came up}
- **Decision:** {what was decided}
- **Implications:** {what this means going forward}
" >> decisions/_active.md

git add decisions/_active.md
git commit -m "decision: {brief title}"
git push

That simple append is safe only for a single writer with a fresh clone. Multi-surface writes need write intents with IDs and timestamps, pull-before-push, and PR-based reconciliation for stale updates or conflicts. The core layer stays human-only; non-active layers should go through review instead of direct overwrite.

Connecting Each Surface (Proposed Integrations)​

Examples only. Some work manually; others need vendor support. The mechanism: read files and inject context.

GitHub Copilot in VS Code​

In your user-level settings.json:

{
  "github.copilot.chat.codeGeneration.instructions": [
    { "file": "~/personal-context/core/boundaries.md" },
    { "file": "~/personal-context/core/communication.md" },
    { "file": "~/personal-context/decisions/_active.md" }
  ]
}

Or reference the repo in any project's custom instructions:

<!-- .github/copilot-instructions.md in any repo -->
For my personal preferences and decisions, reference:
https://github.com/<yourname>/personal-context

Copilot CLI​

Use the current standalone copilot CLI. Put durable CLI instructions in $HOME/.copilot/copilot-instructions.md, then point those instructions at the cloned context repo:

mkdir -p ~/.copilot
cat > ~/.copilot/copilot-instructions.md <<'EOF'
For personal preferences and decisions, read:
- ~/personal-context/core/communication.md
- ~/personal-context/core/boundaries.md
- ~/personal-context/decisions/_active.md

Treat the repo as reference context. Do not rewrite core files without human approval.
EOF

For richer integration, expose the same repo through an MCP server and register it with copilot mcp.

Microsoft Scout​

Scout exposes settings for memory, personality presets, workspace, and permissions. Use those surfaces to mirror the same repo-backed preferences manually or through a sync process. A generated profile file can work as an implementation pattern, but the path below is illustrative, not a documented Scout contract:

# Sync script: pull personal-context → render an illustrative Scout profile
$role = Get-Content ~/personal-context/core/role.md -Raw
$comms = Get-Content ~/personal-context/core/communication.md -Raw
$boundaries = Get-Content ~/personal-context/core/boundaries.md -Raw

@"
# Personal Profile
$role

## Communication
$comms

## Boundaries
$boundaries
"@ | Set-Content ./scout-profile-example.md

Microsoft 365 Copilot​

Sync the repo to a OneDrive folder:

OneDrive/personal-context/ → synced from GitHub repo

Any MCP-Enabled Tool (Claude, Cursor, ChatGPT)​

Expose the repo as an MCP resource server, or clone it locally and point the tool config to the files. MCP is the integration protocol for tools, resources, and prompts; use it to expose context as resources or tools where supported.

Getting Started: Example 30-Minute Setup​

This is a rough first-pass estimate, not a guarantee. The ongoing cost is maintenance: review proposed changes, resolve conflicts, and prune stale active context.

1. Create the repo (5 minutes)​

gh repo create personal-context --private
cd personal-context
mkdir -p core decisions process active .github

2. Write your identity (10 minutes)​

Write what you would tell a new team member on day one.

3. Capture your first decisions (10 minutes)​

Write five repeated preferences or decisions in decisions/_active.md.

4. Connect one surface (5 minutes)​

Wire up VS Code settings, Scout profile, or a CLI alias. Verify it loads context.

5. Evolve naturally​

When an AI asks something it should know, write it down, commit, and push.

The Payoff​

Likely payoff: fewer repeated preferences, fewer re-decisions, and faster starts. Keep time-saved claims only if measured.

Beyond the Repo: When a Service Makes Sense​

The repo is the floor: markdown, git, no server, no vendor, no API key. Use a service only when a flat repo cannot enforce access. Git gives readers the whole file; a hosted service can return only authorized slices.

What a hosted version would buy you​

• Server-side redaction by trust tier. Enforce public / work / private tiers at the server. A flat repo cannot do that; clone access gets everything.
• Identity-based audit and access control. Log who read context, when, and from which surface.
• Central precedence. Resolve boundaries, decisions, process, and active state once instead of per surface.

The key design call: contract vs. transport​

Do not make MCP the canonical contract. MCP is the integration protocol, not the canonical data model. Keep Git as the source of truth. If you build a service, make it a derived read facade over the repo, with a REST/OpenAPI contract and MCP exposed as a thin facade where clients support it.

Keep the contract you can't afford to rewrite in Git and REST; expose MCP as a facade you can afford to replace.

Version the REST facade carefully. Treat MCP adapters as replaceable.

What the service actually is: a context broker​

The service has four jobs:

1. Merge — combine the layers (core, decisions, process, active) into one view.
2. Priority — apply the precedence stack so conflicts resolve deterministically.
3. Redaction — return only the caller's trust tier.
4. Defense-in-depth scanning — flag output that appears to reveal a tier the caller should not see.

The third job is the security boundary for prompt-injection exfiltration. With server-side redaction before retrieval, private context never enters the prompt for an untrusted caller. The fourth job can catch mistakes, but it cannot make unsafe retrieval safe.

Even a service still hits the standards wall​

The limitation: even with a hosted service, consumption stays uneven:

The hosted version still needs a shared standard. For one developer, the repo is usually enough. A service earns its complexity only with multiple trust tiers, multiple consumers, or a real injection threat model.

What's Next: The Standard That Doesn't Exist Yet​

This post is a proposal, not a product announcement. Today, none of this works automatically. Each tool reads its own context files, in its own format, from its own location. Vendors are adding memory, custom instructions, project files, and agent profiles, but not a shared context standard.

What's missing is a shared standard for where personal context lives and how to read/write it. The Model Context Protocol standardized tool integration; user identity needs the same kind of agreement. No shared standard exists today.

The GitHub repo approach is a bet: structured markdown plus a retrieval manifest could work if tool builders agreed to read it.

The ask to tool builders: Add an $AI_CONTEXT_PATH or equivalent. Let users point to markdown context. Portable context works when surfaces agree to read the repo.