Five defects found by auditing the agent stack, all of which let the harness act
confidently on a wrong picture of its own state.
1. A turn that errored discarded the record of tools that had already run. The CLI
hands run_turn a CLONE of the session; the turn appends the user message, each
assistant reply, and a tool_result per executed tool to that clone alone. The error
path dropped it while the side effects of those tools remained on disk, and the
recovery path then replayed the same user input against a session showing no sign
they had ever run — inviting a second execution of tools that are not idempotent.
The clone is now adopted before teardown, and recovery resumes the turn instead of
replaying it (new ConversationRuntime::resume_turn).
2. Every sizing guard was computed against the model NAME's context window. With a
local server behind a cloud model alias that is 200k against a much smaller
backend, which does not make the preflights inaccurate — it makes them unreachable.
CLAW_CONTEXT_WINDOW now declares the real window; model_token_limit honours it and
returns a limit for unrecognised models instead of None, max_tokens is capped to a
quarter of it, and the auto-compaction threshold derives from it (70%) rather than
defaulting to a 200k-window constant.
3. Whether to recover by discarding conversation history was decided by substring
matching on rendered error prose, including strings that ordinary transport and
decode failures also produce — so a dropped connection cost the session its
transcript. Classification now happens in the API layer, where the error still has
its type and the request is still in hand: typed overflow errors, plus ambiguous
transport failures only when the request actually filled the window. The two
copy-pasted detection blocks are gone.
4. The top-level agent loop had no iteration ceiling (usize::MAX) while spawned
subagents were bounded at 32. It now gets the same bound, overridable with
CLAW_MAX_TURN_ITERATIONS.
5. Auto-continue nudges were injected as USER turns, so harness-authored text was
persisted, re-sent on every later request, and folded into compaction summaries as
something the user had said. They are recorded as system turns now; the wire
payload is unchanged.
Also replaces the repository's own self-host section, which described a translator
and a serving stack no longer in the request path. That file is folded into the
system prompt verbatim, so it was briefing the model with an architecture that does
not exist.
(cherry picked from commit c8834cb4d1a3372eb70241429f641162a420297e)
A reply carrying no tool_use ended the turn unconditionally. Small local
models routinely narrate the next tool call in prose — "Now running the
EEC-V probe:" — and then emit EOS, so the announced work never happened
and the user was dropped back at the prompt. The same failure wears a
second hat mid-task: stopping to ask for a go-ahead on work already
requested.
The loop now inspects the final line of a tool-less reply. If it
announces a pending action, or asks a question after tools have already
run this turn, a short nudge is pushed as a user message and the loop
continues instead of breaking. The budget is bounded (default 3 per
turn, CLAW_MAX_STALL_NUDGES, 0 disables) so a model that really is
finished still gets to stop.
Prevention pairs with recovery: the "Doing tasks" system-prompt section
now tells the model to act in the turn it plans in.
Tests cover the auto-continue, the bound, a plain completion that must
NOT be nudged, the disable switch, and last-line-only detection. The
behavioural pair was mutation-checked: reverting the continue to a break
fails both.
(cherry picked from commit 9035fe6f4bf4a8d4cb70d7a3baf1f3748c757498)
The git diff snapshot is rendered into the *system prompt*, so its cost is
paid again by every session whose working tree has changed -- and a diff
changes on essentially every edit. The 50_000-char budget was four times the
entire instruction-file budget (12_000) and unbounded in practical effect: a
single modified file added ~15_000 tokens to the prompt.
Measured against a self-hosted backend, one uncommitted change took the
prompt from 8_302 to 23_457 tokens and a one-word reply from ~3 s to 296 s.
With the new default the same session is 7_911 tokens and 36 s. That cost is
absorbable against a hosted frontier model and ruinous against a local one.
Default is now 4_000 chars, matching MAX_INSTRUCTION_FILE_CHARS, and is
overridable with CLAW_MAX_GIT_DIFF_CHARS following the same pattern as
CLAW_MAX_READ_CHARS. A budget of 0 omits the snapshot entirely rather than
emitting an empty section with a truncation marker; git status and the
recent-commit list are unaffected either way.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RNKqVB2JE9UnCkz8FC9HPy
(cherry picked from commit e01c6ea43d56a97e284ab80fa7a90a479997ed74)
Two related gaps let a single agentic turn exceed a small context window,
with no recovery path once it happened.
1. read_file's character ceiling was a hardcoded 64 KB (~16k tokens), sized
for a 200k window. Two reads in one turn reach ~48k tokens, which a 32k
backend rejects outright. Make it configurable via CLAW_MAX_READ_CHARS,
defaulting to the existing 64 KB so behavior is unchanged unless set.
Reads remain paged, not refused: callers still get truncated/nextOffset.
2. Compaction could not help, because it works at message granularity and an
agentic turn is five messages (user / assistant+tool_use / tool_result /
assistant+tool_use / tool_result). With preserve_recent_messages: 4 the
only droppable message is the short user prompt, so the two oversized
tool results were preserved verbatim by design.
Trim preserved tool-result payloads over 16 KB to a head/tail excerpt with
an explicit "[... N characters elided ...]" marker, so the model knows its
view is partial and can re-read. Small payloads pass through untouched.
Also apply trimming on the early-return path: a session can exceed the
window while holding fewer messages than preserve_recent_messages, where
should_compact is false and message-granular compaction has no move to
make. compact_now no longer gates on removed_message_count alone, which
would have computed the smaller session and then discarded it.
Also fixes a pre-existing flaky test: repl_executes_python_code resolves the
python runtime through PATH, which the PowerShell tests blank while they run.
It now takes the same env_lock, so it no longer intermittently fails with
"python runtime not found" on a machine that has python.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013bVvRYn7CQKR6hT2j1L6Cs
(cherry picked from commit b0d4d41e8132f8dc75717b1f74554d649639de10)
The usage-based compaction gate is post-hoc: it can only see the context
after a call returns and reports its counters. Between that measurement and
the next request the session can grow enormously - a reasoning model emitting
a long thinking block, then a tool result carrying a large file.
Observed live: a call measured at 17,858 prompt tokens generated 10,842 more,
a read_file added ~19k on top, and the next request reached the provider at
47,755 tokens and was rejected outright. No tuning of the post-hoc threshold
prevents that, because nothing measures the session in between.
Add a pre-send check driven by estimate_session_tokens, a local heuristic over
the transcript, so it costs no API call and works before any usage has been
recorded at all.
Note the limit of this: compaction preserves the 4 most recent messages, and
in an agentic tool-use loop those messages ARE the tool results carrying the
bulk. This check bounds gradual growth; it cannot rescue a session whose last
four messages already exceed the window.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013bVvRYn7CQKR6hT2j1L6Cs
(cherry picked from commit 26ab52d70fb17e49b74d523aecfc1da7f69a6d60)
`maybe_auto_compact` gated on `usage_tracker.cumulative_usage().input_tokens`,
which is wrong in two independent ways:
1. It ignores cached input. With prompt caching a warm call reports
`input_tokens: 2` and puts the real context in `cache_read_input_tokens`,
so the counter barely moved and auto-compaction never fired at all,
however large the conversation grew. Backends without prompt caching
report the whole prompt as `input_tokens`, so the bug was invisible there.
2. A sum across calls is not a context measurement, and it never decreases.
Once it did cross the threshold, every subsequent iteration compacted
again, permanently, even after compaction had shrunk the context.
Gate on the latest call's total input instead — input + cache_creation +
cache_read — which is the live context window size. It rises with the real
context and falls back below the threshold once compaction has done its job,
so no counter reset is needed.
`CLAUDE_CODE_AUTO_COMPACT_INPUT_TOKENS` now means what it reads like:
compact when the live context reaches N tokens. The default of 100_000 is
unchanged and still lands at 50% of a 200k window.
Adds `TokenUsage::total_input_tokens()` (saturating) and four tests: the
cached-usage shape that never fired, non-latching across three turns, and
two unit tests for the new helper.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013bVvRYn7CQKR6hT2j1L6Cs
(cherry picked from commit aebba4e77eddf585fc5b21123967dbc4ebb6e6a1)
The suite asserted against whatever the developer's box happened to be
configured as. Two independent leaks, both surfaced by a real `~/.claw`:
1. `parse_args` resolves defaults — notably `permissions.defaultMode` —
through the user-scope config, which `default_config_home()` reads from
`CLAW_CONFIG_HOME` falling back to `$HOME/.claw`. Nothing overrode it, so
a `defaultMode` of `dontAsk` on the host turned 12 assertions from
`WorkspaceWrite` into `DangerFullAccess`. `env_lock()` now returns an
`EnvGuard` that points `CLAW_CONFIG_HOME` at a fresh empty directory and
restores the previous value on drop, so "no user config" is the state
under test. All 42 call sites already bind the guard, so this fixes them
at once; tests that want a populated config home can still set the var
themselves afterwards. `removed_login_and_logout_subcommands_error_helpfully`
asserted a Default-sourced permission mode without taking the guard at
all, and now does.
2. `direct_resume_safe_slash_commands_route_to_local_json_actions_831`
hardcoded `/sandbox` => "warn", which is only true where namespace
isolation is unsupported and just the filesystem sandbox is active
(#731). On a host with working user namespaces the correct answer is
"ok". That entry now accepts either; the test is about #831 routing —
local JSON action instead of `interactive_only` — not about the kernel
under the runner.
Verified with the real ~/.claw/settings.json in place (defaultMode=dontAsk):
cargo test --workspace 0 failures, clippy --workspace --all-targets
-D warnings 0 findings, scripts/fmt.sh clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Q9McEEH32ZtyM62GSXVSyJ
(cherry picked from commit 67aa4530c22f6fdf95a29d6a46992af137134c67)
`cargo clippy --workspace --all-targets -- -D warnings` is this repo's own
documented verification command, and it could not run to completion. Earlier
compile failures were masking a further 48 lints; with those cleared the
workspace now passes with zero findings.
Real fixes, not suppressions:
- status_json_value took 8 positional parameters, four of them a tail of
`Option<&_>` that most callers passed as None — trivially transposable.
Grouped into `StatusJsonExtras` with a Default impl; all 7 call sites now
name only the fields they actually set.
- extract_help_metadata's 8-tuple return is now a documented `HelpMetadata`
alias.
- the auto-compact loop indexed preserve_schedule by its loop counter;
it now iterates the array directly.
- a sandbox status chain had two identical arms; merged the condition.
- a filter_map whose every arm returned Some is now a map.
- 33 unused stdout/stderr bindings in the output-format contract tests are
prefixed with `_`, and 9 `map_or(false, ..)` are now `is_some_and(..)`.
NOTE: several `--bin claw` tests (parse_args defaults) read the *real*
`~/.claw/settings.json` rather than an isolated config home, so they fail on
any machine whose user config sets `permissions.defaultMode`. With that file
moved aside the binary's suite is 207/207. That isolation gap is pre-existing
and left for a separate change.
Verified: clippy workspace/all-targets/-D warnings reports 0; cargo test
--workspace passes with the user config neutralised; scripts/fmt.sh clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Q9McEEH32ZtyM62GSXVSyJ
(cherry picked from commit 03c8a8fd7ba6e405a8217d00d3766c1c6367e527)
`ApiError` is the `Err` half of nearly every `Result` in the api crate, and
its `Api` variant carried ~139 bytes inline (8 fields, four of them
`Option<String>`). That tripped clippy::result_large_err on 35 signatures
and, with -D warnings, made the whole crate unbuildable under the repo's
own documented verification command.
Move the payload into `ApiErrorDetails` behind a `Box`:
- `ApiError::Api(Box<ApiErrorDetails>)`, with `ApiError::api(..)` to
construct and `api_details()` to borrow
- `ApiErrorDetails` re-exported from the crate root
- adding a field here can no longer re-inflate every signature
This also simplifies `enrich_bearer_auth_error`, which previously
destructured and rebuilt all eight fields at five separate return points
just to append a hint; it now mutates the boxed payload in place.
Also clears the remaining pre-existing lints that blocked -D warnings:
unused test bindings, a redundant `use serde_json`, and an explicit
`.into_iter()` in claw-rag-service.
Verified: cargo test --workspace passes, 0 failures.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Q9McEEH32ZtyM62GSXVSyJ
(cherry picked from commit e9210bd75eb76e5d88b2ae797f537102bcabd41d)
An unbounded read_file returned the entire file, which on a small-context
backend overflows the model's window. The backend then rejects the whole
request (llama-server: "request (N tokens) exceeds the available context
size"), which kills the turn — so reading everything at once is precisely
what makes the session stop. Observed live: one 100 KB source file is
~25k tokens against a 32k window with a ~6k-token prefix, and a session
read the same file three times and died on six 400s.
read_file now returns a page and says so:
- absent `limit` yields DEFAULT_READ_LINES (2000) rather than EOF
- MAX_READ_CHARS (64 KB) bounds the payload even when `limit` is
explicit, so minified or generated files can't defeat the line cap
- `truncated` and `nextOffset` tell the caller it has a partial view
and exactly where to resume; the tool description explains both
MAX_READ_SIZE (10 MB) is untouched: it is a process guard against slurping
huge or binary blobs, not a context guard, and at ~2.5M tokens it never
fires on the files that actually overflow a window.
Also clears pre-existing clippy failures in trident.rs that blocked
-D warnings on every edit to this crate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Q9McEEH32ZtyM62GSXVSyJ
(cherry picked from commit c796d7e5eedc66b0cb37bb6823ec772f8ceef9cf)
refactor(cli): let the emitter report the output tail it leaves behind
Two places believed they owned terminal output: consume_stream streams
assistant deltas live while emit_output = true, and run_turn decided
whether to emit a separating newline before spinner.finish by guessing
from final_assistant_text(&summary).
That guess agrees with reality only because every *other* writer happens to
terminate its output with a newline — tool results go through
stream_markdown (which appends one), thinking summaries are written with a
trailing newline, and tool-call headers use writeln! — so the assistant
text path is the only one that can leave a partial line. That is a property
of the current renderers rather than of the streaming protocol, and it is
not enforced anywhere.
TerminalTail is state the emitter records into as it writes: both live
writers (AnthropicRuntimeClient::consume_stream and the tool-result blocks
in CliToolExecutor::execute) now push bytes through a TailTrackingWriter
that reports the last byte written, and the caller that ends a turn asks
where the emitter stopped instead of inferring it.
The heuristic also left the failure arm unguarded. A turn that streamed text
and then errored called spinner.fail on a partial line, which the line
clear then erased; that arm now carries the same guard.
Also gate the animated spinner.tick frame on is_terminal(). That frame is
pure cursor control (save, rewrite line, restore) and emits no newline, so
down a pipe it leaves an unterminated banner that the first streamed delta
then continues on the same physical line.
—
*[repo owner's gaebal-gajae (clawdbot) 🦞]*
- Updated SECURITY.md with explicit instructions for reporting vulnerabilities via GitHub Security Advisories
- Replaced 'when available' qualifier since Private Vulnerability Reporting is now enabled
- Provided step-by-step guidance for security researchers to submit vulnerability reports
Co-authored-by: clawdbot <clawdbot@ultraworkers.dev>
Reviewed and verified: 4 new tests added, all passing. 116 lines added to test JSON escape sequences and i64 boundaries. No production code changes.
—
*[repo owner's gaebal-gajae (clawdbot) 🦞]*
Add Asolaria json=0 wire-format + addressing example
Contributes a reproducible benchmark demonstrating:
- Binary wire format vs JSON text for agent event logs
- 4.64× size reduction (153.2 B → 33.0 B per record)
- ~11× faster encoding
- Tamper-detection via SHA-256 hash-chain
- Self-contained example with no changes to claw-code core
All measurements reproducible; architectural vision clearly distinguished.
Closes#3293
The api.star-history.com SVG endpoint is broken due to GitHub
stargazer API restrictions. Switch to the working alternative at
star-history.dera.page for both dark/light SVGs and the chart link.
Route groq-prefixed models through the OpenAI-compatible client,
apply Groq token-limit metadata, and skip local tool injection for
compound models.
Add regression coverage for routing, token limits, and compound
local-tool behavior.
Co-authored-by: Oz <oz-agent@warp.dev>
docs: add Simplified Chinese translation (README.zh.md)
Adds a complete Chinese translation of the README with proper language selector.
Contribution by @JasonYeYuhe
feat: add persistent auto-memory system
Implement a file-based persistent memory system that allows the AI to retain knowledge about users, feedback, projects, and references across sessions.
## Implementation Details
- MemoryStore data layer with FNV-1a workspace fingerprinting
- MemoryRead/MemoryWrite tools with security hardening (path traversal, symlink, absolute path, and size limit protections)
- System prompt injection of MEMORY.md index with prompt injection mitigation (fenced code block + trust-lowering notice)
- autoMemoryEnabled config option (defaults to true)
- Doctor health check for auto-memory state
- /memory command enhancement showing persistent memory info
## Verification
✓ 1023 lib tests pass
✓ Output format contract tests pass
✓ Release build succeeds
✓ Security hardening verified
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Asserts that an unquoted UNC path like \\server\share\file.txt survives
extract_path_candidates unchanged, preventing the regression where
shlex.split(posix=True) stripped the backslashes.
Replaces the basic extraction test with an end-to-end test that verifies
UNC paths are correctly denied when outside the workspace (e.g. local root),
but correctly allowed when the workspace root is itself a UNC share.
Validates the fix from the PR review.
Adds explicit test coverage to ensure drive-relative paths (e.g. C:foo)
and escaped UNC paths (e.g. \\\\server\\share\\dir) from JSON payloads
are correctly validated by the workspace containment checks.
Closes final review suggestions.
- Updates _check_bash_state to not short-circuit on the first unusable
bash candidate on Windows, ensuring a working fallback is found if present.
- Clears bash detection cache in unittest setUp/tearDown rather than just
relying on a pytest fixture, covering native unittest runners.
Adds a conftest.py pytest fixture to explicitly clear the lru_cache for bash
detection, addressing the monkeypatching hazard in a way that doesn't rely
solely on comments. (Note: CI currently runs python -m unittest, but this
keeps local pytest runners safe.)
Also updates the bash check to run the execution probe directly inside the
cached _check_bash_state, exposing the difference between 'no bash found'
and 'bash present but execution failed' directly to the unittest skip reason.
Moves get_bash_executable and require_bash into a shared module so their
lru_cache state is per-session rather than per-test-module. Also adds a
comment regarding monkeypatching since lru_cache will silently drop changes
to os.environ if it has already been populated.
Closes follow-up items from PR review.
Bellman
dependabot[bot]
DaBear987
reacechalmers
clawdbot
devilking7x
AbdullahFageehandOz
Jesse Daniel Brown
YE
Ming LiuandClaude Opus 4.6
Gracy769