Cursor Agent v2026.07.09-a3815c0 Tip: Use /config to customize Cursor settings and behavior.

give a good detail explanation of every files in this directory and its subdirectories.

Exploring the workspace directory structure and reading key files to produce a detailed explanation of each one.

Globbed "**/*" in .
Found 148 files

$ find /home/ubuntu/.cursor -type f 2>/dev/null | head -500 | sort 1.3s … 186 output lines hidden · ctrl+o to expand /home/ubuntu/.cursor/skills-cursor/update-cursor-settings/SKILL.md /home/ubuntu/.cursor/statsig-cache.json

Reading representative files from each major section to explain their purpose and structure.

Read 8 files
… 5 earlier items hidden
Read .../2da7b294-0888-4db6-bb49-8aff431c9089/meta.json
Read ...0888-4db6-bb49-8aff431c9089/prompt_history.json
Read projects/home-ubuntu-cursor/.workspace-trusted

$ find /home/ubuntu/.cursor -type d | sort && echo “—” && find /home/ubuntu/.cursor -type f | wc -l 420ms … 105 output lines hidden · ctrl+o to expand — 188

Read 4 files
Read ...9089/2da7b294-0888-4db6-bb49-8aff431c9089.jsonl
Read projects/home-ubuntu-cursor/worker.log
Read ...vault-PoC-dotnet-hashicorp/terminals/938586.txt
Read statsig-cache.json
$ sqlite3 /home/ubuntu/.cursor/ai-tracking/ai-code-tracking.db “.schema” 2>/dev/null head -50 530ms 
Grepped, read 1 grep, 4 files
… 2 earlier items hidden
Used sdk
Used create-hook
Read tool output

$ head -c 500 /home/ubuntu/.cursor/statsig-cache.json && echo “” && ls -la /home/ubuntu/.cursor/ai-tracking/ && file /home/ubuntu/.cursor/ai-tracking/ai-code-tracking.db && sqlite3 /home/ubuntu/.cursor/chats/5ffc83748f14b6c02cd2340bde70b055/2da7b294-0888-4db6-bb49-8aff431c9089/store.db “.tables” 2>/dev/null exit 127 • 503ms … 4 output lines hidden · ctrl+o to expand -rw-r–r– 1 ubuntu ubuntu 3399680 Jul 19 02:46 ai-code-tracking.db /home/ubuntu/.cursor/ai-tracking/ai-code-tracking.db: SQLite 3.x database, last written using SQLite version 3044002, file counter 565, database pages 830, 1st free page 203, free pages 64, cookie 0x19, schema 4, UTF-8, version-valid-for 565

Globbed "**/*" in skills-cursor
Found 18 files

$ python3 -c “ import sqlite3 396ms … 10 input + 8 output lines hidden · ctrl+o to expand table: blobs table: meta

Read 13 files
… 10 earlier items hidden
Used review
Used update-cursor-settings
Read projects/home-ubuntu-tthvault/repo.json

This is ~/.cursor — Cursor’s local data directory on your machine. It holds CLI/IDE configuration, synced built-in skills, chat session storage, per-workspace agent state, and analytics. There are 188 files total; many follow repeating patterns (UUID-named sessions, SQLite databases, logs). Below is a full breakdown by area.

────────────────────────────────────────

High-level layout

~/.cursor/ ├── cli-config.json # CLI user settings & auth cache ├── agent-cli-state.json # One-time CLI UI flags ├── statsig-cache.json # Feature-flag cache (Statsig) ├── ai-tracking/ # AI code attribution tracking DB ├── chats/ # Chat UI state (per workspace) ├── projects/ # Agent runtime state (per opened folder) └── skills-cursor/ # Built-in Cursor Agent Skills (synced from server)

┌──────────────────────────────────────────────────────────────────────────────┐ ┌──────────────────────────────────────────────────────┬───────────────────────────────────────────────────────────────────────────────────────────────────────┐ │ Root config │ │ chats/ │ projects/ │ │ │ │ │ │ │ │ │ │ │ │ ┌─────────────────┐ ┌───────────────────────────┐ ┌────────────────────┐ │ ├───────────┐ ┌─────────────────────┐ ┌──────────┐ │ ┌───────────┐ ┌────────────────────┐ ┌────────────┐ ┌─────────────────┐ ┌───────────────────┐ │ ┌───────┐ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ cli-config.json │ │ agent-cli-state.json │ │ statsig-cache.json │ │ │ meta.json │ │ prompt_history.json │ │ store.db │ │ │ repo.json │ │ .workspace-trusted │ │ worker.log │ │ terminals/.txt │ │ agent-tools/.txt │ │ │ chats │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └────────┬────────┘ └───────────────────────────┘ └────────────────────┘ │ ├───────────┘ └─────────────────────┘ └──────────┘ │ └───────────┘ └────────────────────┘ └────────────┘ └─────────────────┘ └───────────────────┘ │ └───┬───┘ │ │ │ │ │ │ │ └──────────┼───────────────────────────────────────────────────────────────────┘ ├──────────────────────────────────────────────────────┘ │ │ │ │ │ │ │ │ │ │ ▼ │ │ │ ┌─────────────────┐ ┌───────────────────────────┐ │ │ │ │ │ │ │ │ │ │ │ projects │ │ agent-transcripts/*.jsonl │◄───────────────────────────┼──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┼─────┘ │ │ │ │ │ │ └─────────────────┘ └───────────────────────────┘ │ │ │ │ └──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘ ctrl+o to show source

────────────────────────────────────────

Root-level files

cli-config.json

Main Cursor CLI configuration. Controls how the agent behaves on the command line.

┌──────────────────────────┬───────────────────────────────────────────────────────────────┐ │ Section │ Purpose │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ permissions.allow / deny │ Which shell commands the agent may run without extra approval │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ editor.vimMode │ Vim keybindings in the CLI │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ display.* │ UI toggles (line numbers, thinking blocks, status line, etc.) │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ model / selectedModel │ Default AI model (e.g. Composer 2.5 Fast) and parameters │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ approvalMode │ How strictly tool use is gated (unrestricted, etc.) │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ sandbox │ Sandbox mode for isolated command execution │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ authInfo │ Cached login identity (email, display name, user ID) │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ privacyCache │ Privacy/ghost mode settings │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ serverConfigCache │ Cached backend URLs (api2.cursor.sh, agent endpoints) │ ├──────────────────────────┼───────────────────────────────────────────────────────────────┤ │ attribution │ Whether commits/PRs are attributed to the agent │ └──────────────────────────┴───────────────────────────────────────────────────────────────┘

This is the file the update-cli-config skill edits.

agent-cli-state.json

Small CLI onboarding / migration flags, not general settings:

• hasShownAgentCommandTip — whether the agent-command tip was shown • hasClearedLegacyStatsigFields — migration flag for old config fields

statsig-cache.json

Large (~537 KB) cached feature flags and experiments from Statsig (https://statsig.com). Cursor uses this to decide which features are enabled (experiments, dynamic configs, etc.) without hitting the network every time. Contains JSON like dynamic_configs, can_record_session, experiment groups, etc. Safe to treat as opaque cache; Cursor refreshes it periodically.

────────────────────────────────────────

ai-tracking/

ai-code-tracking.db

SQLite database for AI code attribution / tracking — which code was AI-generated vs human-written.

Tables:

┌────────────────────────┬────────────────────────────────────────────┐ │ Table │ Likely purpose │ ├────────────────────────┼────────────────────────────────────────────┤ │ ai_code_hashes │ Hashes of AI-generated code chunks │ ├────────────────────────┼────────────────────────────────────────────┤ │ tracked_file_content │ Snapshots/hashes of tracked file content │ ├────────────────────────┼────────────────────────────────────────────┤ │ ai_deleted_files │ Files removed that were previously tracked │ ├────────────────────────┼────────────────────────────────────────────┤ │ scored_commits │ Commits scored for AI contribution │ ├────────────────────────┼────────────────────────────────────────────┤ │ conversation_summaries │ Summaries linked to coding sessions │ ├────────────────────────┼────────────────────────────────────────────┤ │ tracking_state │ Internal sync/cursor state for the tracker │ └────────────────────────┴────────────────────────────────────────────┘

Used for analytics, attribution, and related product features — not something you normally edit by hand.

────────────────────────────────────────

chats/ — chat session storage

Organized as:

chats/// ├── meta.json ├── prompt_history.json ├── store.db ├── store.db-shm (optional, SQLite WAL) └── store.db-wal (optional, SQLite WAL)

The workspace hash (e.g. 5ffc83748f14b6c02cd2340bde70b055) identifies which folder you had open. The session UUID is one chat/conversation.

Workspace folders in your tree

┌──────────────────────────────────┬──────────────────────────────────────────────────────┐ │ Hash folder │ Associated workspace (inferred) │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ 5ffc83748f14b6c02cd2340bde70b055 │ /home/ubuntu/.cursor (current) │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ 7f7140f9ab86684b40aa95373a3bce0a │ /home/ubuntu/tthvault/PoC_dotnet_hashicorp │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ cdd2143e17dbaf8601c198b5900289d6 │ /home/ubuntu/tthvault/PoC_dotnet │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ 96d7b0b89e83553e20e2e4a101bfa5e2 │ /home/ubuntu/tthvault/PoC_go │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ e7a81946e3a063bdc2d9ffe2591cfacc │ /home/ubuntu/tthvault │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ 0f55e2131922b905fb663eebc9baa90d │ (PoC dotnet hashicorp — alternate workspace binding) │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ 8c74903ded871a6ec848c2b39eaca4e3 │ vault-cluster project │ ├──────────────────────────────────┼──────────────────────────────────────────────────────┤ │ dcf1cb88564cab74bbfd120b64688e93 │ partial/abandoned session (meta only) │ └──────────────────────────────────┴──────────────────────────────────────────────────────┘

Per-session files

meta.json — session metadata, e.g.:

{ “schemaVersion”: 1, “createdAtMs”: …, “title”: “Explain All Files”, “cwd”: “/home/ubuntu/.cursor”, “hasConversation”: true, “updatedAtMs”: … }

prompt_history.json — array of user prompt strings for that session (command history / rewind).

store.db — SQLite blob store for the chat UI (blobs, meta tables). Holds message content, attachments, and UI state in a structured way.

store.db-shm / store.db-wal — SQLite Write-Ahead Log sidecar files when a session DB is actively open. Normal; do not delete while Cursor is running.

────────────────────────────────────────

projects/ — agent runtime per workspace

Each subfolder name encodes the workspace path: slashes become hyphens, e.g. home-ubuntu-tthvault-PoC-dotnet-hashicorp → /home/ubuntu/tthvault/PoC_dotnet_hashicorp.

Your project folders

┌─────────────────────────────────────────────────────────┬────────────────────────────┐ │ Folder │ Workspace │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu │ /home/ubuntu │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu-cursor │ /home/ubuntu/.cursor │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu-tthvault │ /home/ubuntu/tthvault │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu-tthvault-PoC-dotnet │ PoC .NET project │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu-tthvault-PoC-dotnet-hashicorp │ PoC .NET + HashiCorp Vault │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu-tthvault-PoC-dotnet-hashicorp-vault-cluster │ Vault cluster variant │ ├─────────────────────────────────────────────────────────┼────────────────────────────┤ │ home-ubuntu-tthvault-PoC-go │ PoC Go project │ └─────────────────────────────────────────────────────────┴────────────────────────────┘

Common files in each project

repo.json — stable UUID for this workspace in Cursor’s indexing/sync system, e.g. {“id”: “4df72b6c-6704-4e37-aab7-e63346182015”}.

.workspace-trusted — records that you approved agent access to this folder:

{ “trustedAt”: “2026-07-19T02:46:38.943Z”, “workspacePath”: “/home/ubuntu/.cursor” }

worker.log — log from Cursor’s background indexing worker for that workspace: Merkle tree sync, file indexing, socket path (worker.sock), errors (e.g. npx not found). Useful for debugging indexing issues.

agent-transcripts//.jsonl — line-delimited JSON transcript of agent turns: user messages, assistant text, tool calls (Glob, Shell, Read, etc.). One file per agent session; this is what agents (and you) use to resume context. Format:

{“role”:”user”,”message”:{“content”:[{“type”:”text”,”text”:”…”}]}} {“role”:”assistant”,”message”:{“content”:[{“type”:”text”,”text”:”…”},{“type”:”tool_use”,”name”:”Shell”,…}]}}

terminals/.txt — snapshot of a shell session the agent started or monitored:

pid: 87012 cwd: “/home/ubuntu/tthvault/PoC_dotnet_hashicorp” command: “docker compose run –rm test 2>&1” started_at: 2026-07-13T09:45:30.822Z running_for_ms: 911214 —

agent-tools/.txt — captured output from long-running tool commands (e.g. full Docker/Vault test logs). Only present when output was too large or needed persistence beyond the terminal snapshot. ──────────────────────────────────────── skills-cursor/ — built-in Agent Skills Synced from Cursor’s servers. Each skill is a folder with SKILL.md (YAML frontmatter + instructions). .sync-manifest.json tracks last sync time per skill. .sync-manifest.json Maps skill name → lastSyncedAt timestamp. Cursor uses this to know when to re-pull skills from the cloud. Individual skills ┌─────────────────────────┬─────────────────────────────────────────────────────────────────────────────────────┐ │ Skill folder │ What it does │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ babysit/ │ Keeps a PR merge-ready: resolve conflicts, triage Bugbot/comments, fix CI in a loop │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ canvas/ │ Builds interactive .canvas.tsx UIs beside the chat (IDE surface) │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ create-hook/ │ Creates Cursor hooks (hooks.json, scripts on agent events) │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ create-rule/ │ Creates Cursor rules (.cursor/rules/, AGENTS.md, coding standards) │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ create-skill/ │ Authoring guide for new SKILL.md files │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ create-subagent/ │ Defines custom subagents (reviewers, debuggers, domain experts) │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ loop/ │ Runs a prompt/skill on an interval (e.g. /loop 5m /foo); disabled in cloud │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ migrate-to-skills/ │ Converts old .mdc rules and slash commands → skills format │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ review/ │ Entry point to choose Bugbot vs Security Review │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ review-bugbot/ │ Runs Bugbot-style code review subagent │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ review-security/ │ Runs security-focused review subagent │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ sdk/ │ Guide for Cursor SDK (@cursor/sdk, cursor-sdk) — programmatic agents │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ shell/ │ Executes text after /shell as a literal shell command │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ split-to-prs/ │ Splits a large change into smaller reviewable PRs │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ statusline/ │ Configures custom CLI status line above the prompt │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ update-cli-config/ │ Edits cli-config.json (permissions, sandbox, model, etc.) │ ├─────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────┤ │ update-cursor-settings/ │ Edits VS Code/Cursor settings.json (themes, format on save, etc.) │ └─────────────────────────┴─────────────────────────────────────────────────────────────────────────────────────┘ Each SKILL.md is markdown with frontmatter (name, description, sometimes disable-model-invocation, metadata.surfaces) plus step-by-step instructions the agent follows when the skill is triggered. ──────────────────────────────────────── How the pieces connect 1. You open a folder → Cursor creates projects// with repo.json, starts worker.log indexing. 2. You start a chat → chats/// stores UI state; projects/.../agent-transcripts/ stores the agent log. 3. The agent runs shell commands → output goes to terminals/*.txt (and sometimes agent-tools/*.txt). 4. Skills in skills-cursor/ are loaded into the agent context when relevant. 5. cli-config.json drives CLI behavior; ai-tracking/ and statsig-cache.json support product/analytics features. ──────────────────────────────────────── What you can safely touch vs leave alone ┌────────────────────────────────────────────────────────────────────┬──────────────────────────────────────────────────────┐ │ Safe to edit (with care) │ Leave alone (managed by Cursor) │ ├────────────────────────────────────────────────────────────────────┼──────────────────────────────────────────────────────┤ │ cli-config.json (via CLI or skill) │ statsig-cache.json │ ├────────────────────────────────────────────────────────────────────┼──────────────────────────────────────────────────────┤ │ Custom skills you add under ~/.cursor/skills/ (not skills-cursor/) │ store.db, ai-code-tracking.db │ ├────────────────────────────────────────────────────────────────────┼──────────────────────────────────────────────────────┤ │ Rules in your project’s .cursor/rules/ │ store.db-wal, store.db-shm while running │ ├────────────────────────────────────────────────────────────────────┼──────────────────────────────────────────────────────┤ │ │ Most of chats/ and agent-transcripts/ (session data) │ └────────────────────────────────────────────────────────────────────┴──────────────────────────────────────────────────────┘ Note: skills-cursor/ is Cursor-managed — edits may be overwritten on sync. User-authored skills typically live in ~/.cursor/skills/ (not present in this tree yet). ──────────────────────────────────────── Summary by file type (for the ~188 files) ┌──────────────────────┬─────────────────┬────────────────────────────────┐ │ File pattern │ Count (approx.) │ Role │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ SKILL.md │ 17 │ Built-in skill definitions │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ *.jsonl │ ~25 │ Agent conversation transcripts │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ meta.json │ ~30 │ Chat session metadata │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ prompt_history.json │ ~28 │ User prompt history per chat │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ store.db (+ WAL/SHM) │ ~30+ │ Chat blob storage │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ terminals/*.txt │ ~20 │ Shell session captures │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ worker.log │ 8 │ Per-workspace indexer logs │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ repo.json │ 8 │ Workspace IDs │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ .workspace-trusted │ 5 │ Trust approvals │ ├──────────────────────┼─────────────────┼────────────────────────────────┤ │ Root JSON + 1 DB │ 4 │ Global config & tracking │ └──────────────────────┴─────────────────┴────────────────────────────────┘ If you want, I can go deeper on one area — for example every session under PoC_dotnet_hashicorp, or the full schema inside store.db / ai-code-tracking.db.