review: M5d-1 epochs and compaction

Co-Authored-By: Grok 4.6 <noreply@x.ai>

docs/implementer-lessons.md

@@ -33,7 +33,8 @@ How it is used:
 | I15 | Never write the model's chat-template tags out whole (the think and tool-call tags, the angle-bar turn markers) in code, comments, commands or the log. Build them from pieces (`concat!("<", "think")`) and name them in words. Text holding them turns into the model's own control tokens when it reads it. | M5a task 04: the given test held the tags; Ornith's reasoning ended at the test file's closing think tag, and it lost track of the file and made no change. `docs/research/2026-09-24-chat-template-tokens.md`. | yes | M5a tasks 05 to 22: not seen again. |
 | I16 | When a test outside your task fails, check once whether it fails without your change (stash, run, unstash), then stop and report both results. Do not keep experimenting, and never add a file the task does not list, not even a temporary one. | M5a task 21: the session did the check, then spent about 80 minutes on experiments and wrote two diagnostic test files into `crates/bxctl/tests/` until the owner stopped the run; the cause was outside the repository (review finding 4). M5a task 16 did it right: one check, its log row, a stop. | yes | M5c-1: the stash left the new untracked tests behind and a baseline printed no total (task 15; now `git stash -u` in AGENTS.md, T53); the two flaky tests were rerun and named, not stopped on (now an exception for a test the README lists, T52). |
 | I17 | Write your log row from `git diff` after staging, naming only what the diff shows. A row written from memory describes what you meant to do. | M5a task 10's row describes checks that are not in `serve.rs`; task 13's says two runbook fragments were copied, and neither was (review findings 5 and 6 of 2026-09-24). | yes | M5c-1: task 10 logged one gate run and ran three, task 12 logged three and ran four and "Deviations: none" after anchoring an edit on its own line, and task 14's row blames the task for its own spacing error. M5c-2: every row's first gate is `pass`; task 01 still ran the gate twice, and every session ran `cargo test --workspace` after `gate: ok` only to count (T59). |
-| I18 | After `gate: ok`, do not run `cargo test --workspace` (or the gate) only to count tests. The gate's total is the count. If HEAD already has this task's commit and a `done` row, stop; do not start the task again. | M5c-2: every session re-ran the workspace tests after a passing gate (T50 still driving extra work). Task 05's run log is a resume after HEAD was already at task 12; Laguna did not re-commit, and the original transcript was truncated. | yes | |
+| I18 | After `gate: ok`, do not run `cargo test --workspace` (or the gate) only to count tests. The gate's total is the count. If HEAD already has this task's commit and a `done` row, stop; do not start the task again. | M5c-2: every session re-ran the workspace tests after a passing gate (T50 still driving extra work). Task 05's run log is a resume after HEAD was already at task 12; Laguna did not re-commit, and the original transcript was truncated. | yes | M5d-1: tasks 03, 07 and 08 ran the gate two or three times with first-gate pass. |
+| I19 | Add the log row immediately under the table header (newest first). Do not insert it by date among older rows. | M5d-1 task 06: the row was inserted after M4b/03. | yes | |
 
 ## Tips for writing tasks
 
@@ -99,6 +100,7 @@ How it is used:
 | T58 | For a class ceiling or trust rule in a spec, list every path that writes to the egress it guards (turns, mirrored memory writes, undo, review accept, notes) and give each path a test at the forbidden class. A rule written for one path is not checked on the others by default. The same for a rule that a class *must* still take a path: give that path a test at that class. | M5c-1 review finding 1: the `secret` ceiling was built and tested for turns only; memory writes from a `secret` session were mirrored to the trusted and fleet workspaces (the reference had it too). Finding 2 is the same shape for undo. M5c-2 review finding 1: a `secret` session's Honcho query was implemented and unit-tested for `has_secret`, but not at the serve layer against the fake Honcho. |
 | T59 | After `gate: ok`, the task is done counting. Do not tell the implementer to run `cargo test --workspace` (or the gate) again only to compare a number. AGENTS.md already forbids a second gate "only to count"; say the same of the workspace test run. | M5c-2: all twelve sessions ran `cargo test --workspace` after a passing gate to rediscover the total, even with T50's wording in the task. |
 | T60 | The driver must not truncate a task's run log, and must not start a task whose commit and `done` row are already in HEAD. A power-loss resume continues the existing log file; it does not open it for write from offset zero, and it skips a task that is already committed. | M5c-2 task 05: `.state/runs/M5c2/05-memoryd-honcho-readers.log` is four `=== …` resume sections after HEAD was at task 12; the original transcript is gone. Commit `b370c60` was already complete. |
+| T61 | Name where the log row goes: immediately under the header, newest first. The implementer otherwise inserts by date among older rows. | M5d-1 task 06. |
 
 ## What worked and should be kept
 
@@ -161,3 +163,13 @@ difference an equivalent rewrite. Its friction was the plan's: an undefined test
 forbidden `git switch` in every task (T51), and reaching for a crate's source the task did not
 give (T56). The milestone's serious defect was the spec's reach, not the code's: a class ceiling
 applied to turns and not to memory writes (T58), in the reference too.
+
+M5c-2 was Ornith through the driver, twelve tasks, first-gate pass in every row, no stop. Three
+skeleton bodies were equivalent rewrites of the reference; the rest matched. Extra workspace test
+runs after `gate: ok` (I18) and a truncated power-loss resume (T60) were process only. The one
+product finding was a missing serve-layer test of a path the code already took (T58).
+
+M5d-1 was Ornith through the driver, eight tasks, first-gate pass in every row, no stop. The only
+code difference from the reference is an equivalent rewrite in `compact.rs`. Extra gate runs after
+a pass (I18) and a log row inserted among M4b (I19) were process only. The device check passed
+spec 7 on the second run, after a trusted identity write then a tainted recall.

docs/implementer-log.md

@@ -8,6 +8,7 @@ reviewer adds findings under "Reviews" once per milestone.
 |---|---|---|---|---|---|---|---|
 | M5d1/08-loopd-compact-turn | 2026-10-03 | done | 3 | pass | none | Copied the given test `crates/loopd/tests/compact_turn.rs` (new). Step 2 (`cargo build --tests --keep-going -p loopd`) showed exactly the named failure: `compact_turn` not compiling with E0425 (`cannot find function finish_turn in module compact`) — the task named E0599, this is E0425 because it is a free function not a method; every other target compiled, including `turn` and `channel`. Made F1 (`compact.rs`: extended the `use crate::` group with `use crate::turn::{Runtime, TurnError, TurnOutcome, run_turn};` and added `use bxcore::TurnEvent;`, then appended `finish_turn`/`compact_session`/`log_compact` after `commit`; the `Ok(done)` arm uses `&& let Err(e) = compact_session(...)` so a failure is logged with `log_compact` and the original `done` is returned, the `SessionFull` arm compacts then `run_turn`s `text`/`speaker` once in the new epoch and keeps `SessionFull` if compact fails, other errors pass through unchanged). Made H1 (`channel.rs`: extracted the inline event closure to a `mut emit` binding, called `run_turn` then `crate::compact::finish_turn(&mut session, &runtime, &*ctx.memory, &turn.content, turn.speaker.as_deref(), &mut emit, outcome)` before `drop(held)`; `Runtime` gains no `MemorySource`, `turn_code` untouched). `cargo check -p loopd` clean, no error or warning. `wc -l` prints 186 for `compact.rs` and 284 for `channel.rs`, each ≤ 500. `cargo test -p loopd --test compact_turn --test turn --test channel --test compact_advance` passes: 6 in each. `git diff --stat` lists only `compact.rs` and `channel.rs`; `git status --short` shows `compact_turn.rs` as new. `cargo fmt --all` changed nothing after the edits. `make gate` printed `gate: ok` with 1508 tests (1502 before, plus the 6 of `compact_turn.rs`). No new dependency, so `Cargo.lock` unchanged. | ornith-1.5-35b-a3b |
 | M5d1/07-loopd-compact-advance | 2026-10-03 | done | 2 | pass | none | Copied the given test `crates/loopd/tests/compact_advance.rs` (new). Step 2 (`cargo build --tests --keep-going -p loopd`) showed exactly the named failure: `compact_advance` with E0425 (`cannot find function commit in module compact`); every other target compiled. Made S1 (`session.rs`: `begin_epoch` after `log_path`, in that order — writes `<n>.baseline.json` with `to_json`, `create_new` the `<n>.jsonl`, accumulates `prior_users`/`prior_untrusted` from the old `records` before clearing them, sets `epoch`/`baseline`/`carry`/empty `records`/the new `File`, then `persist` `SessionStart` with the live `slot`; does not reset `next_call` or `slot`). Made C1 (`compact.rs`: extended the `use crate::` group to `Baseline`/`MemorySource`/`Registry` and the bxcore group to `Epoch`/`Timestamp`, then added `commit` after `summarise` — skips the append and takes `next`/summary from a trailing `EpochEnd`, else appends `EpochEnd` with `Epoch(epoch.0.saturating_add(1))`, reads `memory.block()` as `Failed`, `Baseline::assemble_with(cfg, registry, &block)` as `Failed`, then `begin_epoch`). `cargo check -p loopd` clean, no error or warning. `wc -l crates/loopd/src/session.rs` prints 438 and `crates/loopd/src/compact.rs` prints 134, each ≤ 500. `cargo test -p loopd --test compact_advance --test session_epochs --test compact_summary` passes: 6 in `compact_advance`, 9 in `session_epochs`, 6 in `compact_summary`. `git diff --stat` lists only `session.rs` and `compact.rs`; `git status --short` shows `compact_advance.rs` as new. `make gate` printed `gate: ok` with 1502 tests (1496 before, plus the 6 of `compact_advance.rs`). No new dependency, so `Cargo.lock` unchanged. | ornith-1.5-35b-a3b |
+| M5d1/06-loopd-compact-summary | 2026-10-03 | done | 1 | pass | none | Edited `compact.rs` E1 (module doc, added the `use`s for `neutralise`, `messages_with_carry`, `Config`, `ChatMessage`/`ChatRequest`/`Client`, `is_context_full`, and the `CompactError` enum plus its `Display`/`Error` impls whose `Display` ends with the `session-compact-failed` pointer) and E2 (added `summarise`: read `compact.md` with the path in the error, build the background-slot request with empty `tools` and thinking on, call `chat_with_retry`, neutralise-and-cap the completion, fall back only on a context-full 400, else `Failed`). Copied `tests/compact_summary.rs`, appended the runbook's `session-compact-failed` entry. `compact.rs` is 92 lines; `cargo check` clean, all 12 compact tests pass, `make gate` prints `gate: ok`, 1496 tests. | Ornith-1.5-35B-A3B |
 | M5d1/05-loopd-compact-due | 2026-10-03 | done | 1 | pass | none | Copied the given test `crates/loopd/tests/compact_due.rs` and the starter `config/compact.md` from `docs/plans/M5d1/files/`. Created `crates/loopd/src/compact.rs`: `due` does `cache_n.saturating_add(prompt_n) as f64 >= threshold * (n_ctx as f64)`; `fallback_summary` `filter_map`s `LogRecord::User { content, .. }` to `&str`, reverses, joins `"\n\n"`, then `cap_chars`; `cap_chars` is the given `chars().take(n)`. Added `pub mod compact;` after `pub mod channel;` in `lib.rs` (rustfmt order, accepted). `cargo check -p loopd` clean, no error or warning. `cargo test -p loopd --test compact_due` passes all 6. `git diff --stat` lists only `lib.rs` and `compact.rs`; `git status --short` shows `compact_due.rs` and `compact.md` as new. `make gate` printed `gate: ok` with 1490 tests (1484 before, plus the 6 of `compact_due.rs`). No new dependency, so `Cargo.lock` unchanged. | ornith-1.5-35b-a3b |
 | M5d1/04-memoryd-honcho-epochs | 2026-10-02 | done | 1 | pass | none | Copied the given test `crates/memoryd/tests/honcho_epochs.rs` (new). Step 2 (`cargo build --tests --keep-going -p memoryd`) showed exactly the named failure: `honcho_epochs` with one E0432 (`unresolved import memoryd::honcho::turns::read_session`); every other target compiled. Made H1 (`turns.rs`: `log_path` doc now names the sibling epoch files and that `read_session` concatenates them), H1b (inserted `epoch_logs` after `read_records`: lists `<n>.jsonl` under `sessions/<id>/`, `n` parsed as `u32`, sorted by `n`, a missing directory empty; then `read_session` concatenates `read_records` over each), H2 (`resume` reads `read_session` instead of `read_log(&log_path(...))`), H3 (`records` matches `read_session(...).map_err(|e| e.to_string())`, keeping the logged line shape); W1 (`worker.rs`: added `use ...decide::{final_reply, turn_records, user_text}` and `use ...turns::{LogError, TurnSync, log_path, read_session}`), W2 (`send_turn` loads `read_session`, `self.readable(&path)` on Ok, `LogError::Damaged(_)` -> `Unsent::Local("log")`, other error -> `Unsent::Unreadable(e.to_string())`, then `turn_records`/`user_text`/`final_reply` from `decide`), W3 (`batch` called with `text`/`reply` by value, no borrow). `read_turn` no longer used in the worker body but stays re-exported. `cargo check -p memoryd` clean, no error or warning. `wc -l` prints 369 for `turns.rs` and 453 for `worker.rs`, each ≤ 500. `cargo test -p memoryd --test honcho_epochs --test honcho_turns --test honcho_worker` passes: 6 in `honcho_epochs`, 16 in `honcho_turns`, 12 in `honcho_worker`. `git diff --stat` lists only `turns.rs` and `worker.rs`; `cargo fmt --all` changed nothing after the edits. `make gate` printed `gate: ok` with 1484 tests (1478 before, plus the 6 of `honcho_epochs.rs`). No new dependency, so `Cargo.lock` unchanged. | ornith-1.5-35b-a3b |
 | M5d1/03-loopd-carry-messages | 2026-10-02 | done | 2 | pass | none | Copied the given test `crates/loopd/tests/carry_messages.rs` (new). Step 2 (`cargo build --tests --keep-going -p loopd`) showed exactly the named failure: `carry_messages` with one E0432 (`unresolved import loopd::baseline::{CARRY_HEADING, messages_with_carry}`); every other target compiled. Made B1 (`baseline.rs`: `CARRY_HEADING` = `concat!("Earlier ", "turns, summarised:")`, `messages` becomes `messages_with_carry(baseline, records, None)`, new `messages_with_carry` inserts the heading plus a blank line plus the summary as one `ChatMessage::User` after the system prompt and before walking `records`; the rest of the `match` untouched); T1 (`turn.rs`: `use crate::baseline::messages_with_carry`); T2 (`run_turn` passes `session.carry()`). `cargo check -p loopd` clean, no error or warning. `cargo test -p loopd --test carry_messages --test baseline --test turn` passes: 5 in `carry_messages`, 6 in `baseline`, 6 in `turn`. `git diff --stat` lists only `baseline.rs` and `turn.rs`; `cargo fmt --all` changed nothing. `make gate` printed `gate: ok` with 1478 tests (1473 before, plus the 5 of `carry_messages.rs`). No new dependency, so `Cargo.lock` unchanged. | ornith-1.5-35b-a3b |
@@ -94,7 +95,6 @@ reviewer adds findings under "Reviews" once per milestone.
 | M4b/05-brokerd-config-gateway | 2026-09-24 | done | 1 | pass | none | Copied `docs/plans/M4b/files/crates/brokerd/tests/config_gateway.rs` and the new `support/rig.rs` (whose `Config` gained `gateway: None`) into `crates/brokerd/tests/`. In `config.rs`: added the `Gateway { socket: PathBuf }` struct (`#[serde(deny_unknown_fields, default)]`) before `default_podman`, the `gateway: Option<Gateway>` field (with `#[serde(default)]`) after `runner` in `Config`, and the `gateway_socket()` method before `audit_dir` — `None` when the section is absent, the default `<home>/run/gateway/gateway.sock` when the socket is empty, else the path as given. `cargo fmt --all` then `cargo test -p brokerd` green with `config_gateway` 4 passed; `make gate` prints `gate: ok` on the first run. | Ornith-1.5-35B-A3B |
 | M4b/04-brokerd-answer-post | 2026-09-24 | done | 1 | pass | none | Moved `enum Answer` and `struct Answered` from `ledger.rs` into `approvals.rs` before `pub struct Entry` (with doc comments), added `pub use crate::approvals::{Answer, Answered};` next to the `Verdict` use and `DecisionRecord` to the `use proto` line in `approvals.rs`. Added the `post` field to `Answer::Approved`/`Answer::Refused`, threaded it through `ledger::answer` (step 3 binds `post`, step 6 writes it into `AuditEvent::Approval` instead of `None`), and gave the two `bxctl` answers in `admin.rs` `post: None`. Added `Table::contains` before `take`. Copied `ledger_post.rs`, `broker_pending.rs`, `broker_sequence.rs` and `ledger_answer.rs` into `crates/brokerd/tests/`; every suite passes, `ledger_post` 3 passed, `ledger.rs` is 487 lines, and `make gate` prints `gate: ok` on the first run. | Ornith-1.5-35B-A3B |
 | M4b/03-bxctl-deny | 2026-09-24 | done | 1 | pass | none | Copied `docs/plans/M4b/files/crates/bxctl/tests/cli_deny.rs` into `crates/bxctl/tests/`. In `parse`, `"refuse" => parse_refuse(rest, home)` became `"refuse" | "deny" => parse_refuse(rest, home)`, so `deny` produces the same `Command::Refuse` as `refuse` for every argument list including the error cases. Added a blank line and the two-line `bxctl deny <id> [--reason TEXT] [--admin-socket PATH]` / `Another name for refuse.` block to `USAGE` after the `refuse` block, keeping the two-space and six-space indents. The copied test's `deny_parses_as_refuse` and `usage_names_deny` both failed before the changes and pass after; `cargo test -p bxctl` is green and `make gate` prints `gate: ok` on the first run. | Ornith-1.5-35B-A3B |
-| M5d1/06-loopd-compact-summary | 2026-10-03 | done | 1 | pass | none | Edited `compact.rs` E1 (module doc, added the `use`s for `neutralise`, `messages_with_carry`, `Config`, `ChatMessage`/`ChatRequest`/`Client`, `is_context_full`, and the `CompactError` enum plus its `Display`/`Error` impls whose `Display` ends with the `session-compact-failed` pointer) and E2 (added `summarise`: read `compact.md` with the path in the error, build the background-slot request with empty `tools` and thinking on, call `chat_with_retry`, neutralise-and-cap the completion, fall back only on a context-full 400, else `Failed`). Copied `tests/compact_summary.rs`, appended the runbook's `session-compact-failed` entry. `compact.rs` is 92 lines; `cargo check` clean, all 12 compact tests pass, `make gate` prints `gate: ok`, 1496 tests. | Ornith-1.5-35B-A3B |
 | M4b/02-proto-gateway-wire | 2026-09-24 | done | 2 | fail | none | Copied `gateway.rs`, `gateway_wire.rs` and the eight `gateway_*.json` fixtures into place; added `pub mod gateway;` after `pub mod frame;` and the re-export to `lib.rs`, the four `Message` variants plus the `use` to `wire.rs`, and the four `kind` arms (changing "fourteen" to "eighteen") in `broker.rs`. Filled `is_mattermost_id` (`id.len() == 26 && every byte a-z0-9`) and `ApprovalReply::check` in the given order: `by`, `post`, then a reason over `MAX_REASON` chars by `chars().count()`. All 3 `gateway_wire` tests pass; the first `make gate` failed on clippy `collapsible_if`, fixed by collapsing the nested `if let` into a let-chain guard, and the second prints `gate: ok`. | Ornith-1.5-35B-A3B |
 | M4b/01-proto-escape | 2026-09-24 | done | 1 | pass | none | Copied `docs/plans/M4b/files/crates/proto/src/escape.rs`, `bxctl/src/escape.rs` and `proto/tests/escape.rs` into place, and added `pub mod escape;` to `proto/src/lib.rs` after `pub mod class;`. The bxctl file is now the one-line re-export `pub use proto::escape::{escape_json_text, escape_model_text};`, so no bxctl source or test changed and behaviour is identical. `cargo test -p proto --test escape` reports 8 passed; every bxctl suite passes unchanged (the moved escape test now runs against `proto::escape`); `cargo fmt --all` then `make gate` prints `gate: ok` on the first run. | Ornith-1.5-35B-A3B |
 | M4a/16-gatewayd-secret-file-race | 2026-09-24 | done | 2 | pass | Reworded the template comment in `read_checked` to drop the literal `check_file` token, so step 5's grep prints only the credential branch's `std::fs::read(&path)` | Replaced `check_file` with `pub fn read_checked(path, between) -> Result<Zeroizing<Vec<u8>>, String>`: check the path (absolute; `symlink_metadata` as `named`; not a symlink; a regular file), call `between()`, open the path, take `opened` metadata, refuse if `(opened.dev(), opened.ino()) != (named.dev(), named.ino())`, then the owner-uid and mode checks on `opened` (not `named`), then read the whole file into a `Zeroizing<Vec<u8>>` from that same handle. Added `use std::io::Read;`. In `load`, the File branch now calls `read_checked(path, &|| {})` and `value(std::mem::take(&mut *bytes))` so nothing unwiped remains. The three suites pass (5, 8, 5); the only remaining `std::fs::read(` is the credential branch and `check_file` is gone. `make gate` prints `gate: ok`. | Ornith-1.5-35B-A3B |
@@ -1003,3 +1003,53 @@ markup (`ask_memory` stays at `minimal`); turn numbers restart per epoch; a `bxc
 `sanitize::refuse`'s secret rules to replace `has_secret`; Mattermost usernames `a.b` and `a_b`
 map to the same Honcho peer; queries from `secret` sessions go to Honcho on the assumption in
 spec 8 and `docs/egress.md`.
+
+### M5d-1, tasks 01 to 08 — reviewed 2026-10-03 by the design model (Grok 4.6)
+
+Accepted. Tasks 01 to 08 by Ornith-1.5-35B-A3B through `tools/run-plan.sh` on this machine,
+device-checked on straylight the same day. The branch differs from the reference
+(`~/boxmaker-refs/m5d1-ref.bundle`, `e05845b`) in `crates/loopd/src/compact.rs` only: an equivalent
+rewrite of `due` and `fallback_summary`, and a split of the import group. `docs/implementer-log.md`
+differs because of the log rows. Every other product file matches.
+
+| Check | Result |
+|---|---|
+| 8 task commits, each with `Co-Authored-By: ornith-1.5-35b-a3b <agent@wntrmute.dev>` | pass |
+| Given tests byte-identical to `docs/plans/M5d1/files/` (the runbook fragment is appended into `docs/runbook.md`) | pass |
+| Protected files (`docs/design.md`, `docs/decisions.md`, `docs/inference-contract.md`, `docs/specs/`, `docs/plans/`, `AGENTS.md`, `CLAUDE.md`) untouched by Ornith | pass |
+| Log rows | 8, all `done`; first gate pass in every row. Extra gate runs on 03 (2), 07 (2), 08 (3). Task 06's row was inserted after M4b/03; moved under the header in this review. Model cell on 06 is `Ornith-1.5-35B-A3B`; the trailer is `ornith-1.5-35b-a3b`. |
+| No `unwrap`, `expect`, `panic!`, `#[allow]` or `unsafe` in new library code; every source file ≤ 500 lines; no new crate | pass |
+| `make gate` on straylight (rustc 1.98.1 / clippy 0.1.98) | `gate: ok`, 1508 tests, the plan's count |
+| Independent review of `git diff master m5d1 -- crates` against spec sections 3, 5, 7 and 8 | no product finding; `priors` walks `0..n`; carry is `ChatMessage::User` with `CARRY_HEADING` from `concat!`; summary on the background slot, empty tools, thinking on; `finish_turn` before `drop(held)`; Honcho concatenates every `<n>.jsonl`; `session-compact-failed` has a pointer and a runbook entry |
+
+| # | Severity | Where | Finding | Resolution |
+|---|---|---|---|---|
+| 1 | process | implementer (tasks 03, 07, 08) | Extra `make gate` runs after a first-gate pass, only to count. | I18 seen again |
+| 2 | process | implementer (task 06) | Log row inserted by date among M4b instead of immediately under the header. Model cell does not match the trailer. | Row relocated in this review; I19 / T61; AGENTS.md now says where the row goes |
+
+Device check, 2026-10-03, on straylight (`target/debug` of `m5d1` at `b274fa7`; Ornith slot 0 live,
+slot 2 background; slot 1 left busy; a private `$HOME`, production daemons untouched). `[compact]
+threshold = 0.0` so every turn compacted. Empty `grants/` (memory tools are built-in). First run
+failed the baseline check: the model recalled a planted tainted note on turn 1, tainting the
+session, then wrote the teapot fact with `action: note` into `notes.md` marked untrusted, so
+`memory.md` stayed empty and `1.baseline.json` lacked the fact. Compact, epoch-0 recall, and
+`untrusted` persistence still worked. Second run wrote identity first (`memory` `add` `file=memory`),
+then the tainted recall, then the epoch-0 phrase. Every item of the spec's section 7 for M5d-1
+passed:
+
+| Check | Result |
+|---|---|
+| Next baseline contains the `memory.md` fact after compact | Turn 1 stored `M5d1-check: the cobalt teapot lives in the attic.` and replied `Stored. magenta-walrus-epoch0`; `1.baseline.json` contains the fact; `TurnDone.untrusted` false |
+| `recall` of an epoch-0 phrase still hits | Turn 3: `magenta-walrus-epoch0` hit the session log |
+| `TurnDone.untrusted` stays true across the epoch change after a tainted recall | Turn 2 recalled `TAINTED-COBALT…`; `untrusted` true on turns 2 and 3; epochs 0–3 on disk |
+
+What was good: eight tasks, first-gate pass in every row, no stop, no product bug against the
+reference or the spec. Compaction, carry, Honcho concatenation and taint spanning epochs all held
+on the live model.
+
+Open items carried forward from M5c-2, with "turn numbers restart per epoch" closed by this half:
+rift's dialectic at `reasoning: low` still returns raw tool-call markup (`ask_memory` stays at
+`minimal`); a `bxcore` function for `sanitize::refuse`'s secret rules to replace `has_secret`;
+Mattermost usernames `a.b` and `a_b` map to the same Honcho peer; queries from `secret` sessions
+go to Honcho on the assumption in spec 8 and `docs/egress.md`. Known M5d-1 limits, left for
+M5d-2: live turns still use `slots.main`; `[compact] summarise` is unused; no idle compact path.