Replaces PR #3493's blanket fatal abort with a "tell the model + throttle
the bypass loop" policy. Workspace-bound rejections are now ordinary
recoverable tool errors enriched with a structured "this is a hard policy
boundary" instruction; SSRF stays the only marker that aborts the turn.
Why the fatal-abort approach broke
----------------------------------
PR #3493 promoted every shell `_guard_command` and filesystem path-resolution
rejection to a turn-fatal RuntimeError. Two of those messages (`path
outside working dir` and `path traversal detected`) are heuristic substring
scans on the raw command, so legitimate commands like `rm <ws>/x.txt
2>/dev/null` or `find . -type f` killed the user's turn (#3599). On
channels with outbound dedupe (Telegram) the user just saw silence (#3605),
and the noise polluted the LLM's context until it started hallucinating
guard rejections on plain relative paths (#3597).
Why we still need *some* throttle
---------------------------------
The original #3493 pain point was real: the LLM, refused once, would
swap tools and try again -- read_file -> exec cat -> exec cp -> bash -c
-> ln -sf -> python -c open(...). Just removing the fatal escape lets
that loop run wild until max_iterations.
What this commit does
---------------------
- `nanobot/utils/runtime.py`: add `workspace_violation_signature` and
`repeated_workspace_violation_error`. The signature normalizes
filesystem `path` arguments and the first absolute path inside an
exec command, so swapping tools against the same outside target hits
the same throttle bucket. Two soft attempts are allowed; the third
attempt's tool result is replaced with a hard "stop trying to bypass"
message that quotes the target path and tells the model to ask the
user for help.
- `nanobot/agent/runner.py`: split classification into `_is_ssrf_violation`
(still fatal) and `_is_workspace_violation` (now soft). All three
failure branches in `_run_tool` (prep_error / exception / Error
result) route through a shared `_classify_violation` that bumps the
per-turn workspace_violation_counts dict and either keeps the tool's
own message or substitutes the throttle escalation. `_execute_tools`
now threads that dict alongside the existing external_lookup_counts.
- `nanobot/agent/tools/shell.py`: append a structured boundary note to
every workspace-bound guard rejection (`working_dir could not be
resolved`, `working_dir is outside`, `path outside working dir`,
`path traversal detected`). SSRF errors stay short and direct so the
model doesn't try to "phrase around" them. Existing `2>/dev/null`
allow-list and benign device passthrough from the previous commit
remain.
- `nanobot/agent/tools/filesystem.py`: append the same boundary note to
the `outside allowed directory` PermissionError so read_file / write_file
/ list_dir errors give the LLM the same explicit hint.
Tests
-----
- `tests/utils/test_workspace_violation_throttle.py` (new): signature
collapses across read_file/exec/python -c against the same path,
different paths get independent budgets, escalation only fires after
the third attempt.
- `tests/agent/test_runner.py`:
- `test_runner_does_not_abort_on_workspace_violation_anymore` -- v2
contract: filesystem PermissionError is now soft, runner moves to
the next iteration and finalizes cleanly.
- `test_is_ssrf_violation_remains_fatal` + the existing
`test_runner_aborts_on_ssrf_violation` -- SSRF still aborts on the
first attempt.
- `test_runner_lets_llm_recover_from_shell_guard_path_outside` -- end
to end recovery from `path outside working dir`.
- `test_runner_throttles_repeated_workspace_bypass_attempts` -- four
bypass attempts against the same outside target produce at least
one `workspace_violation_escalated` event and the run completes
naturally without aborting the turn.
- The two `_execute_tools` direct-call tests now pass the new
workspace_violation_counts dict.
- `tests/tools/test_tool_validation.py`: relax three `==` assertions
to `startswith` + "hard policy boundary" substring check to match
the new structured error messages.
- `tests/tools/test_exec_security.py` keeps the prior `2>/dev/null`
regression and the `> /etc/issue` negative case from the previous
commit on this branch -- they still pass under the new policy.
Coverage status: full pytest 2648 passed / 2 skipped (was 2638 / 2
on origin/main). Ruff is clean for every file touched in this commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
PR #3493 promoted every shell `_guard_command` rejection to a turn-fatal
RuntimeError. The two heuristic outputs in that list -- `path outside
working dir` and `path traversal detected` -- routinely false-positive on
benign constructs (e.g. `2>/dev/null`, quoted `..` arguments to sed/find,
absolute paths inside inline scripts), so legitimate workspace commands
silently kill the user's turn (#3599) and the agent never gets a chance
to retry with a different approach (#3605).
Two changes, both narrowly scoped:
- `ExecTool._guard_command` now skips a small allow-list of kernel device
files (`/dev/null`, the standard streams, `/dev/random`, `/dev/fd/N`,
...) before the workspace path check, matched against the pre-resolve
string so symlinks like `/dev/stderr -> /proc/self/fd/2` still hit the
allow-list. Real outside writes such as `> /etc/issue` remain blocked.
- `AgentRunner._WORKSPACE_BLOCK_MARKERS` keeps only the four hard
path-resolution errors from filesystem.py / shell.py and the SSRF
marker. The two heuristic substrings move out of the fatal list, so
the LLM sees them as ordinary tool errors and can self-correct in the
next iteration. SSRF stays fatal because retrying an internal URL
with a different phrasing would defeat the safety boundary.
Tests:
- `tests/tools/test_exec_security.py`: parametrized regression for the
exact #3599 command sample plus other stdio redirects and device
reads; explicit negative case asserts `> /etc/issue` is still blocked.
- `tests/agent/test_runner.py`: `_is_workspace_violation` no longer
fatals on the two heuristic markers, plus an end-to-end case proving
the runner hands the guard error back to the LLM and finalizes the
next turn cleanly.
* fix: allow_patterns take priority over deny_patterns in ExecTool
Previously deny_patterns were checked first with no bypass, meaning
allow_patterns could never exempt commands from the built-in deny list.
This made it impossible to whitelist destructive commands for specific
directories (e.g. build/cleanup tasks).
Changes:
- shell.py: check allow_patterns first; if matched, skip deny check
- shell.py: deny_patterns now appends to built-in list (not replaces)
- schema.py: add allow_patterns/deny_patterns to ExecToolConfig
- loop.py/subagent.py: pass allow_patterns/deny_patterns to ExecTool
- Add test_exec_allow_patterns.py covering priority semantics
* fix: separate deny pattern errors from workspace violation detection
The deny pattern error message "Command blocked by safety guard" was
included in _WORKSPACE_BLOCK_MARKERS, causing deny_pattern blocks to be
misclassified as fatal workspace violations. This meant LLMs had no
chance to retry with a different command — the turn was aborted
immediately.
Changes:
- shell.py: deny/allowlist error messages now use distinct phrasing
("blocked by deny pattern filter" / "blocked by allowlist filter")
- runner.py: remove "blocked by safety guard" from
_WORKSPACE_BLOCK_MARKERS so deny_pattern errors are treated as normal
tool errors (LLM can retry) instead of fatal violations
- workspace path errors still use "blocked by safety guard" and remain
fatal as intended
* fix: update test assertions to match new deny pattern error message
* fix: indentation error in test file
* fix: restore SSRF fatal classification and tidy exec pattern plumbing
Address review feedback on the deny/allow_patterns rework:
- runner.py: re-add "internal/private url detected" to
_WORKSPACE_BLOCK_MARKERS. The earlier marker removal also stripped
fatal classification from SSRF / internal-URL rejections (whose
message still says "blocked by safety guard"), turning a hard
security boundary into something the LLM could retry.
- loop.py / subagent.py: drop `or None` between ExecToolConfig and
ExecTool. The schema default is an empty list and ExecTool already
normalizes None back to [], so the indirection was a no-op.
- shell.py: extract `explicitly_allowed` flag in _guard_command so
allow_patterns are scanned once instead of twice and the control
flow no longer relies on a no-op `pass + else` branch.
- tests/agent/test_runner.py: add a regression test asserting that
the SSRF block message is treated as fatal, while deny/allowlist
filter messages are deliberately non-fatal.
* fix: remove unused exec allow-pattern test import
Keep the new ExecTool allow-pattern coverage clean under ruff.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Xubin Ren <xubinrencs@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Treat workspace and safety guard failures as fatal regardless of whether they arrive from tool preparation, returned tool output, or raised exceptions.
Made-with: Cursor
MCP resource/prompt/tool names containing spaces or special characters
(e.g. "PostgreSQL System Information") were forwarded verbatim to model
provider APIs, causing validation errors from both Anthropic and OpenAI
which require names matching ^[a-zA-Z0-9_-]{1,128}$.
Add _sanitize_name() that replaces invalid characters with underscores
and collapses consecutive underscores. Applied in MCPToolWrapper,
MCPResourceWrapper, MCPPromptWrapper constructors and the enabled_tools
filtering logic.
Closes#3468
Add two focused regression tests for the retry-wait leak this PR fixes:
- tests/agent/test_runner.py::test_runner_binds_on_retry_wait_to_retry_callback_not_progress
locks in that `AgentRunSpec.retry_wait_callback` (not `progress_callback`) is
what `_build_request_kwargs` forwards to the provider as `on_retry_wait`.
- tests/channels/test_channel_manager_delta_coalescing.py::TestRetryWaitFiltering
runs `_dispatch_outbound` end-to-end and asserts that `_retry_wait: True`
messages never reach channel send.
Both tests fail on origin/main and pass with this PR's fix applied.
Made-with: Cursor
- Extract synthetic user message string to module-level constant
- Tighten comments in _snip_history recovery branch
- Strengthen no-user edge case test to verify safety net interaction
When _snip_history truncates the message history and the only user message
ends up outside the kept window, providers like GLM reject the resulting
system→assistant sequence with error 1214 ("messages 参数非法").
Two-layer fix:
1. _snip_history now walks backwards through non_system messages to recover
the nearest user message when none exists in the kept window.
2. _enforce_role_alternation inserts a synthetic user message
"(conversation continued)" when the first non-system message is a bare
assistant (no tool_calls), serving as a safety net for any edge cases
that slip through.
Co-authored-by: darlingbud <darlingbud@users.noreply.github.com>
Add a built-in tool that lets the agent inspect and modify its own
runtime state (model, iterations, context window, etc.).
Key features:
- inspect: view current config, usage stats, and subagent status
- modify: adjust parameters at runtime (protected by type/range validation)
- Subagent observability: inspect running subagent tasks (phase,
iteration, tool events, errors) — subagents are no longer a black box
- Watchdog corrects out-of-bounds values on each iteration
- Enabled by default in read-only mode (self_modify: false)
- All changes are in-memory only; restart restores defaults
- Comprehensive test suite (90 tests)
Includes a self-awareness skill (always-on) with progressive disclosure:
SKILL.md for core rules, references/examples.md for detailed scenarios.
Keep late follow-up injections observable when they are drained during max-iteration shutdown so loop-level response suppression still makes the right decision.
Made-with: Cursor
- Migrate "after tools" inline drain to use _try_drain_injections,
completing the refactoring (all 6 drain sites now use the helper).
- Move checkpoint emission into _try_drain_injections via optional
iteration parameter, eliminating the leaky split between helper
and caller for the final-response path.
- Extract _make_injection_callback() test helper to replace 7
identical inject_cb function bodies.
- Add test_injection_cycle_cap_on_error_path to verify the cycle
cap is enforced on error exit paths.
When the agent runner exits due to LLM error, tool error, empty response,
or max_iterations, it breaks out of the iteration loop without draining
the pending injection queue. This causes leftover messages to be
re-published as independent inbound messages, resulting in duplicate or
confusing replies to the user.
Extract the injection drain logic into a `_try_drain_injections` helper
and call it before each break in the error/edge-case paths. If injections
are found, continue the loop instead of breaking. For max_iterations
(where the loop is exhausted), drain injections to prevent re-publish
without continuing.
* feat(agent): add mid-turn message injection for responsive follow-ups
Allow user messages sent during an active agent turn to be injected
into the running LLM context instead of being queued behind a
per-session lock. Inspired by Claude Code's mid-turn queue drain
mechanism (query.ts:1547-1643).
Key design decisions:
- Messages are injected as natural user messages between iterations,
no tool cancellation or special system prompt needed
- Two drain checkpoints: after tool execution and after final LLM
response ("last-mile" to prevent dropping late arrivals)
- Bounded by MAX_INJECTION_CYCLES (5) to prevent consuming the
iteration budget on rapid follow-ups
- had_injections flag bypasses _sent_in_turn suppression so follow-up
responses are always delivered
Closes#1609
* fix(agent): harden mid-turn injection with streaming fix, bounded queue, and message safety
- Fix streaming protocol violation: Checkpoint 2 now checks for injections
BEFORE calling on_stream_end, passing resuming=True when injections found
so streaming channels (Feishu) don't prematurely finalize the card
- Bound pending queue to maxsize=20 with QueueFull handling
- Add warning log when injection batch exceeds _MAX_INJECTIONS_PER_TURN
- Re-publish leftover queue messages to bus in _dispatch finally block to
prevent silent message loss on early exit (max_iterations, tool_error, cancel)
- Fix PEP 8 blank line before dataclass and logger.info indentation
- Add 12 new tests covering drain, checkpoints, cycle cap, queue routing,
cleanup, and leftover re-publish
Keep tool-call assistant messages valid across provider sanitization and avoid trailing user-only history after model errors. This prevents follow-up requests from sending broken tool chains back to the gateway.
- Adjusted message handling in AgentRunner to ensure that historical messages remain unchanged during context governance.
- Introduced tests to verify that backfill operations do not alter the saved message boundary, maintaining the integrity of the conversation history.
- Merged latest main (no conflicts)
- Added test_llm_error_not_appended_to_session_messages: verifies error
content stays out of session messages
- Added test_streamed_flag_not_set_on_llm_error: verifies _streamed is
not set when LLM returns an error, so ChannelManager delivers it
Made-with: Cursor
When the LLM returns an error (e.g. 429 quota exceeded, stream timeout),
streaming channels silently drop the error message because `_streamed=True`
is set in metadata even though no content was actually streamed.
This change:
- Skips setting `_streamed` when stop_reason is "error", so error messages
go through the normal channel.send() path and reach the user
- Stops appending error content to session history, preventing error
messages from polluting subsequent conversation context
- Exposes stop_reason from _run_agent_loop to enable the above check
