v0.3.47

37 commits since v0.3.46. 52 files changed, +2665 / -289 lines.

✨ Features

• e3f20a64 Proactive tool discovery: agent searches for tools before claiming inability, plus project-directive loading
• c8ddb521 Confidential file protection: SSH keys, .env, credentials protected. Owner verification required in group chats
• 87621927 Per-project brain overlay: project-specific brain files layer on top of profile brain files
• 1adc8b17 Owner impersonation detection: detect non-owners trying to act as the owner in Telegram group chats
• 0ddd8479 bot_owner config field: TelegramConfig with is_owner() helper for identity checks
• 67b55edf /profiles command: manage AI profiles from any channel
• a63b042c /cd hidden dirs toggle: show/hide hidden directories in the directory picker

🔧 Fixes

• 217a431e Forum topic session isolation: each forum topic gets its own session (#215)
• b24f13a9 Follow-up question topic routing: questions go to the correct topic thread, not #general
• ee367d2d System brain rebuild on change: system brain rebuilt from disk when brain files change (#213)
• 56b8b057 JIT tool activation: extended tools activated on-demand when called by name (#214)
• 56df0352 tool_search guidance: extended tool guidance moved to preamble + RSI (#214)
• 5d0b4b47 RSI tool success rate: pre-execution misses no longer penalize success rate (#214)
• 53620509 Internal-state query routing: queries route to their tools, not the raw DB
• c00b4a83 State query routing refined: further cleanup of internal-state query routing
• cc689374 Profile/project directives: default profile stated, profile and project directives added
• 6889668d Session recovery hint: corrected to use plan operation="start"
• a37820cb Compaction reinjection fix: TOOLS.md and CODE.md no longer re-injected after compaction
• c6f900ce Plan auto-approve: plan auto-approved on first start
• 238ded51 rm-blocklist bypasses closed: reversed flags, quoted $HOME, long flags, chained rm
• 3dbd430e Blocklist recursion: recurse blocklist through interpreter indirection
• 1328f8fa @botName stripping defense: defense-in-depth stripping + tracing for group command bugs
• e4aa995b TOOLS.md template cap: trimmed below 100-line regression cap

📖 Documentation

• 8036b7e3 Require a tool_search reminder when naming extended tools
• f225cc5c Add YAML frontmatter requirement for SKILL.md files

🧹 Miscellaneous

• 6f762022 Move AGENTS.md to last injection position
• 16c53b56 Drop routing-map duplication from AGENTS.md Owns header
• 17075a23 Drop redundant contextual-files pointer; AGENTS.md owns it
• 2d77b301 Move operational how-tos out of AGENTS into their owners
• 0070d118 Gate build/test on cargo-audit so advisories fail fast
• 09a83340 Live system-brain rebuild test
• 0d522443 Accept STT-chain umbrella error so CI isn't network-fragile
• 0ef60dd3 Per-project brain overlay integration test
• 64b349d2 Docker e2e adversarial eval harness
• 425f1713 cargo fmt commands.rs
• bbd4247d cargo fmt for commands and telegram
• 8ced3bc7 Bump quinn-proto 0.11.14 -> 0.11.15 (RUSTSEC-2026-0185)

📊 Stats

• 37 commits since v0.3.46
• 52 files changed, +2665 / -289 lines
• 4,103 tests (4,103 passed, 0 failed, 24 ignored)

Full Changelog: v0.3.46...v0.3.47