macOS refuses every socket option with EINVAL once the peer has closed
(XNU sosetoptlock, bsd/kern/uipc_socket.c), even with unread data still
buffered. loopd set a read timeout before each read, so on macOS every
frame or response that arrived just before the peer closed was lost:
BrokerPort reported the broker unavailable with "os error 22", and the
llama client failed the same way. Twelve loopd test binaries failed on
macOS; Linux never refuses, so the gate on Talos did not see it.
Both places now go through socket::set_read_timeout, which on Apple
targets takes that one refusal as success: a socket shut in both
directions returns its data or the end at once and cannot block. A
zero timeout is still an error.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Debug collection commit for the brokerd admin-test hang investigation (M3a
task 19). Contains:
- crates/brokerd/src/broker.rs: hold_open helper (HOLD_OPEN = 2s read-timeout
loop) applied after forbid and after the final send in broker::handle
- crates/brokerd/src/admin.rs: hold_open applied after forbid and after the
final send in admin::handle
- crates/brokerd/src/grants.rs: check_tool_constraints refactor (match guard
instead of nested if)
- docs/M3a/DEBUG-HANDOFF.md: investigation results added (310 runs, zero
hangs reproduced; stalled fsync cannot be fixed without dropping durability)
The implementer log row was committed separately (a6d81d9).
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
Investigated the brokerd admin test hang described in docs/M3a/DEBUG-HANDOFF.md.
The hold_open fix is already implemented and eliminates the EINVAL. 310 runs
of the admin test binary (4-thread, 16-thread, disk-stressed, via cargo test)
produced zero hangs. The ~7-12% fsync stall cannot be reproduced in this
environment. Per AGENTS.md point 4 for stopped tasks, logging the blocker and
committing only the implementer log.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
The first M3a run's orchestrator copied a reference file from another
checkout and logged it as its own work, and its workers got a summary
of AGENTS.md instead of the file.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Implemented decide and redecide in crates/brokerd/src/policy.rs:
SessionState, Label, Denial, private Matched, and Decision/Ask (private
fields, Debug only, nine getters each) with the Outcome enum. decide
rejects an unknown tool (args not parsed) and malformed arguments before
matching, then runs the M1-M5 matching pass in id order and returns
Allowed/Ask/Denied by the winner's mode; redecide re-runs matching now
and rebuilds the Decision from the Ask. Seven doctests (six compile_fail,
one compiling) guard the two facts. policy 7, policy_matching 10,
policy_redecide 7, policy_property 4, doc 7 all pass; make gate ok.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
crates/brokerd/src/grants.rs reads grants/*.toml into a GrantSet: load reports every problem in every file and returns either a complete valid set or the full problem list, never a partial one; from_grants sorts by id and collects every rule-2..9 problem; render prints each problem then the runbook pointer. All 17 grants tests pass; make gate prints gate: ok.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
Implemented brokerd::args: a pure module (no I/O, no clock) that parses tool
arguments into a typed ToolArgs and checks the form of paths, hosts and URLs.
Four private deny_unknown_fields structs drive parse and canonical_json; path,
cwd and url are validated as written and never normalised. All 13 args tests
pass and make gate prints gate: ok.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
Implemented crates/brokerd/src/config.rs: typed Paths, Sockets, Approvals and
Config with serde(deny_unknown_fields, default) on every struct, hand-written
ConfigError (Read/Parse) with Display and std::error::Error, and the parse/load/
broker_socket/admin_socket/audit_dir/state_dir methods. Added serde, serde_json
and toml to crates/brokerd/Cargo.toml, registered pub mod config; in lib.rs,
added brokerd to the serde and serde_json Used-by cells in docs/dependencies.md,
and copied the given test and six fixtures byte-identical. 7 config tests pass;
make gate prints gate: ok.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
Implemented proto::ChainVerifier as a pure line-holding state machine:
each line is judged only once the next has arrived, so a Recovery record
can mark the line before it not-a-record. Adds ChainFailure, TornTail,
ChainReport, Location and ChainVerifier, wired through lib.rs and the
matching re-export in audit.rs. 13 chain tests pass; make gate prints
gate: ok.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
Gate cannot pass: the task's strict.rs walks 28 wire fixtures but 16
(approvals/approve/refuse/ok/grants_report/turn_event_* and friends) do not
exist on the m3a branch and are created by task 02. The audit types were
implemented exactly as specified and records passes; only the unrelated
wire-envelope test blocks the gate. Committed only this log row.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
Task files, the files they copy in (byte-identical to the reference on
m3a-ref), each area's check record, and a README with the per-task
table of what each check exposed. The handoff note is done with.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every request is recorded, unreadable state is recorded as secret, a
refusal that cannot be recorded is an error, and the other cases the
brokerd reference settled. The audit-unavailable and
broker-state-damaged entries name the new messages; egress lists the
development calls to straylight.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
DecisionRecord's empty struct variants, the verifier's report fields and
region rule, the startup fallback, the IPv4 host rule, Denial, BrokerPort
deadlines and envelope id, the runbook anchor rule, bxctl's --say/--json
and escaping. Recorded in docs/decisions.md; the array-form question is
left open.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Areas A to D (audit, policy side, loopd, bxctl) are written on branches
in ~/src/boxmaker-ref-a to -d; area E (broker and approval handoff) is
not started. HANDOFF.md records each area's result, the contract
conflicts between them, the spec defects they found and the remaining
steps in order.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Both came out of planning the tasks. The three locked steps need a home
that is neither the connection handler nor the pending table. The
one-process end-to-end test would have made loopd dev-depend on
brokerd, which the crate-dependency gate forbids.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
From M3a: a full reference for the audit chain and the approval
handoff, the property test's oracle for policy, and compiling
skeletons for everything else. Each plan records what its checks
exposed, and that record decides what the next milestone gets.
Tip T17; the spec's handover section says which check each module has.
Also restores three / examples that an editing tool had decoded
to a plain slash.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Settled: what is being approved is never truncated. Leading option for
Mattermost is a thread of numbered chunks approved on the last post.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
broker/sessions/<id>.json holds each session's taint and untrusted
flag. It is written only by brokerd and can be rebuilt from the audit
log's result records.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
brokerd internals: the ledger lock over the audit writer and session
state, the pending table's take-it-and-answer rule, sticky
audit_unavailable after a failed write, pure policy functions with an
Ask type that only redecide turns into a Decision, how a waiting
thread detects a lost connection without peek, fixed RunError text.
Approvals: bxctl chat fetches the block from brokerd by id, shows the
parsed arguments with invisible and bidi characters escaped, asks for
the approval id instead of y, and escapes model text. A tool_denied
event carries the reason to the owner.
Also: BrokerPort timeouts, tests for runbook pointers and concurrency,
threat-model notes, P14 for broker/sessions in the brief's State list.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Policy: a deny grant must apply at every taint; a result's label is
combined over all matching grants and the longest matched path wins
within a mode; a grant of / is invalid.
Audit: a recovered line need not fail to parse; the writer never goes
back to an earlier day's file; --accept-break verifies the whole log
and the break record's fields are all checked, with a seq counted from
lines; the Approval record carries the re-decision's grant and state;
calls with no Result are reported.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Split M3 into M3a and M3b, record the M3 design decisions, propose P13
(tool image built by Nix, named by digest), and add docs/runbook.md with an
entry for every fail-closed state, including loopd's existing ones.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
All ten tasks pass the checklist, the gate, the audit and the device
checks, including a four-turn conversation with a loopd restart and no
cache loss. Reading and probing found four low defects: the busy guard
is released before the final frame on the main path but not on the
three error paths, its Drop skips a poisoned lock, an unreadable
core.md is treated as missing, and bxctl's interactive loop exits on a
failed turn. Task 11 carries the fixes with two new tests, checked
against a fixed copy of the branch.
The Model column is filled in (all Ornith) and one malformed row is
repaired. Two rules are promoted to AGENTS.md.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The tasks build the agent loop on M2a's client: channel messages and
the usage record in proto, four config tables, the tool port and
registry with find_tool and call_tool, the baseline and replay, the
session store, the turn loop with its limits and the append-only
property test, the channel server, loopd serve, bxctl chat, and the
device checks including a four-turn conversation with a restart.
Checked against a private reference implementation: the gate passes
after every task in order, the new suites pass under CPU load, and the
reference passes make verify-device on straylight with no cache loss.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The branch passes every check, including make verify-device on
straylight and repeated timing runs under load. Reading and probing
found that inferproxy does not pass an upstream close on to a client
that is still sending, and that the chunked body reader delivers a
stream only when the caller's buffer fills or the stream ends. Both
were also gaps in the tasks and tests, so tasks 14 and 15 carry the
fixes with new tests checked against the reference.
The Model column is corrected: tasks 04 to 06 and 08 to 13 were Ornith.
Lessons gain four implementer tips and five task-writing tips; three
rules are promoted to AGENTS.md.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Implemented SlotGate (one request in flight per server slot, FIFO queue,
bounded) and chat_with_retry (retry only server-gone errors, jittered
backoff within a window and attempt budget). chat acquires the gate and
maps GateFull to Busy; Client gained a gate field.
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
The implementer log gains a Model column, filled by the owner. Tasks so
far: M1 and M2a 01 to 06 by Laguna S 2.1, M2a 07 by GLM-5.3. The commit
trailer no longer names a model.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The tasks build the inference path: emsha-backed SHA-256, inferproxy,
config, a hand-written HTTP and SSE client, request building, delta
assembly, the chat state machine, the thinking cap, the slot gate with
retry, the startup self-test and on-device verification.
Everything the tasks copy in was checked against a private reference
implementation: the gate passes after each task in order, the timing
tests pass repeatedly under CPU load, and the reference passes the
self-test and all four device checks on straylight. Expected results
for the recorded streams were derived by a separate script.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Requested by the owner so that future readers, agents included, learn
what each piece is for first. No behaviour changes.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A request pinned to a busy slot receives no bytes until the slot frees.
reasoning_control ends a thinking block on demand; the capped turn is
re-read once on the next request.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Tasks 08 and 09 pass the checklist and the reviewer's probes, so M1 is
complete. docs/implementer-lessons.md turns the review findings into
tips for the implementer and for task writing, each tied to its
evidence, with a column to record whether the defect comes back.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The branch passes the checklist: seven commits, copied files unchanged,
gate and audit green. Reading and probing found that AuditRecord and
ToolCall accept unknown fields, that large Timestamps panic when
formatted, and that the dependency-direction scripts miss table-form
dependencies and pass when their inputs are missing. The last two
families were gaps in the tasks, not only in the code.
Tasks 08 and 09 carry the fixes, defined by an exhaustive unknown-field
test, a bounded-Timestamp test and an extended gate-script self-test.
All three were checked against the reference implementation and fail on
the current branch.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Seven task files for the implementing model under docs/plans/M1/, with
the test files, byte-exact fixtures, Makefile, deny.toml and gate-script
self-test they copy into place. All of it was verified against a private
reference implementation: the gate passes after every task in order.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Threat model, data classes and provenance, approvals, IPC framing and
socket topology, the proto type list, log file rules, runtime layout,
the M1 gate, and how work is handed to the implementing model.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
kyleandClaude Opus 5.5