Per-agent file ownership with Unix permission enforcement:
- Agent directories owned by agent ID (UID), group 'agent' (GID)
- Private files (plan, ctl, fifo): mode 0600 - owner only
- Group-readable (chat, log): mode 0440 - owner + agent group
- World-readable (state, id): mode 0444 - observable by all
- Prompt: mode 0220 - CLI and owner can write
virtfs: fix UID/GID inheritance through nested paths
- Added findChildWithInheritance() to accumulate inherited UID/GID
- Stat now correctly shows agent ID as owner for nested files
server: admin bypass for server owner
- serverAdmin variable captures the Unix user running olliesrv
- Admin bypass includes empty uname, 'admin', or server owner
Documentation updates:
- fs/doc.go: 'The Namespace IS the Security Model'
- registry/doc.go: capability-based tool access
- peer.go: capability-based peer access
- lessons-learned.md: 'Model compliance is not a security boundary'
- architecture-9p.md: per-agent file ownership section
Security evaluation:
- Added experiments/security-eval/ with NERV attack corpus adaptation
- Test scripts for Landlock sandbox validation
- RESULTS.md documenting 0% ASR on hostile operations
This implements the NERV thesis: 'An agent can only access resources
explicitly bound into its namespace.' Enforcement is structural via
file permissions, not behavioral via model compliance.
The sudo credential broker was never functional and added complexity
without value. This removes:
- Sudo field from bypass Request structs (broker, client, toolsrv)
- Sudo parameter from EvaluateRequest interface and implementations
- Sudo/ResetsCounter fields from MetaFile and Variant structs
- sudo: true from system_logs.meta variants
- All sudo documentation from writing-tools.md, tool-registry.md,
core.md, evolution.md, and misc.md
Bypass remains fully functional for sandbox escapes. Tools that need
elevated privileges should handle that internally or be run manually.
- Rename 'chat' to 'log': returns last 64KB sliding window (static read).
Tail still works via Qid.Vers. Full history persisted to disk per turn.
- Add 'chat' streaming file: blocking read that delivers new output as
the agent produces it. Per-fid offset tracked via waitBase. Blocks
indefinitely between turns (no EOF). EOF only on kill/session death.
- Add LongBlock interface to File: streaming files bypass the 5s read
timeout. Server returns content directly (ignores Tread offset for
streaming files since waitBase tracks position).
- Add 'kill'/'.'' ctl aliases with FIXME: currently kills entire session.
When multi-agent-per-session lands, kill should kill agent only.
- Rewrite acme frontend (cmd/Ollie) to use 9P client directly instead
of FUSE mount. streamChat() is a blocking read loop — no polling.
All file operations (read, write, ls) go through plan9/client.
- Add 9P streaming prototype in experiments/9p-stream demonstrating
the blocking-read pattern for token delivery.
The key insight: 9P's request-response model gives natural streaming.
Server holds the Tread until data arrives, client blocks on Read().
No polling, no signals, no offset tracking needed.
Updated the documentation to clarify issues encountered during the agent orchestration process, including notification failures, state desynchronization, and hallucination of unnecessary files. Added proposed recovery procedures and future work suggestions to enhance agent performance.
Levi Neely
Ollie Agent
Levi Neely
Levi Neely