docs: Add Studio documentation and UTF-8 safety invariants

README.md:
- Add Studio section documenting the multi-agent workspace manager
- Document usage: run, list, status, accept, discard commands
- Explain worktree-based isolation and workflow

AGENTS.md:
- Add UTF-8 safe string slicing as critical invariant (#8)
- Add MUST NOT for byte-index slicing on multi-byte text (#5)
- Document parser sanitization as dangerous/subtle code path
  (prevents parser poisoning from inline tool-call JSON patterns)

Agent: lamport

AGENTS.md

@@ -14,6 +14,7 @@
 5. **File paths in tools support tilde expansion** - `~` expands to home directory
 6. **Streaming is preferred** - Non-streaming requests block UI
 7. **Tool results are size-limited** - Large outputs are truncated or thinned automatically
+8. **String slicing must be UTF-8 safe** - Use `chars().take(n)` or `char_indices()`, never byte slicing like `&s[..n]` on user-facing strings
 
 ### MUST NOT Do
 
@@ -21,6 +22,7 @@
 2. **Never store secrets in logs** - API keys are redacted in error logs
 3. **Never modify files outside working directory without explicit permission**
 4. **Never assume tool results fit in context** - Large results are thinned automatically
+5. **Never use byte-index string slicing on text with potential multi-byte characters** - Causes panics on emoji, CJK, box-drawing chars
 
 ## Recommended Entry Points
 
@@ -71,6 +73,13 @@
 - Different configs for interactive vs autonomous mode
 - **Risk**: Aggressive retries can hit rate limits harder
 
+### Parser Sanitization (`g3-core/src/streaming_parser.rs`)
+
+- Sanitizes inline tool-call JSON patterns to prevent parser poisoning
+- Replaces `{` with fullwidth `｛` (U+FF5B) when patterns appear inline (not on their own line)
+- Real tool calls from LLMs always appear on their own line
+- **Risk**: Inline JSON examples in prose can trigger false tool call detection without sanitization
+
 ## Do's and Don'ts for Automated Changes
 
 ### Do

README.md

@@ -322,6 +322,44 @@ G3 automatically saves session logs for each interaction in the `.g3/sessions/`
 
 The `.g3/` directory is created automatically on first use and is excluded from version control.
 
+## Studio - Multi-Agent Workspace Manager
+
+Studio is a companion tool for managing multiple g3 agent sessions using git worktrees. Each session runs in an isolated worktree with its own branch, allowing multiple agents to work on the same codebase without conflicts.
+
+### Usage
+
+```bash
+# Build studio alongside g3
+cargo build --release
+
+# Run an agent session (creates worktree, runs g3, tails output)
+studio run --agent carmack "fix the memory leak in cache.rs"
+
+# Run a one-shot session without a specific agent
+studio run "add unit tests for the parser module"
+
+# List all sessions
+studio list
+
+# Check session status (shows summary when complete)
+studio status <session-id>
+
+# Accept a session: merge changes to main and cleanup
+studio accept <session-id>
+
+# Discard a session: delete without merging
+studio discard <session-id>
+```
+
+### How It Works
+
+1. **Isolation**: Each session creates a git worktree at `.worktrees/sessions/<agent>/<session-id>/`
+2. **Branching**: Sessions run on branches named `sessions/<agent>/<session-id>`
+3. **Tracking**: Session metadata is stored in `.worktrees/.sessions/`
+4. **Workflow**: Run → Review → Accept (merge) or Discard (delete)
+
+Studio is the recommended way to run multiple agents in parallel on the same codebase, replacing the deprecated flock mode.
+
 ## Documentation Map
 
 Detailed documentation is available in the `docs/` directory: