Follow-up to train 22: scanner quoting, logger recall, multi-root diagnostics, docs

CHANGELOG.md

@@ -262,6 +262,37 @@ per page instead. Gate: `test/impactdepthcheck.sh`.
 placeholder changes). Every ingest cache written by an earlier build is refused and re-indexed once, and every
 cached quality snapshot is recomputed. The session legend dictionary is `dictv=0e543e1e6a3fe37d entries=750`.
 
+### Fixed — `--scan-skill(s)`: quoting a sensitive path no longer hides its upload (sensitive-read-upload)
+
+`cat /etc/passwd | curl … @- URL` grades CRITICAL, but `cat "/etc/passwd" | curl …` scanned clean: an
+opening quote put the flow scan back in command position, so the reader's quoted file was read as a command and its
+read was never noted. A quote no longer re-arms command position inside a reader's segment (it still does elsewhere,
+so `sh -c "…"` and `echo "…" | sh` are read as code), and a shell word split by quotes or escapes (`/etc/"passwd"`,
+`@"/etc/passwd"`, `/etc/pass\wd`, `< /etc/"passwd"`) is also read whole, once, when it ends — and in command position
+names the command (`\cat`, `c"a"t`, `\curl`). A runner option's split value stays a value (`sudo -u ro''ot cat`), a split upload
+option still marks its file (`curl -""T /etc/passwd`), and a reader's state stays inside its own command context
+(`cat <(sh -c 'curl … @/etc/passwd …')` is an upload). Single-quoted, mixed, redirected and `head -c "N" FILE` forms are
+covered by `test/skillscan.sh`. Against the previous scanner no finding is lost on the skill fixtures, an adversarial
+set or this repository's own Markdown and scripts; two CRITICAL rows gain `why="sensitive-read-upload"`.
+
+### Fixed — `--cochange` / `--owners` / `--hotspots` in a multi-root workspace name a shallow secondary root when they refuse
+
+When no workspace root had a commit to mine, the refusal probed only the primary root for a shallow clone, so a full
+primary next to a depth-limited secondary still said "git unavailable / no history". The refusal now probes every
+mined root (the one root `--owners=SYM` mines) and names the shallow root it found, as the success paths' `shallow="1"`
+already did. Single-root output is unchanged. Gate: `test/shallowhistorycheck.sh` section 3m.
+
+### Fixed — `--quality-delta` error-masking: log-only recognises loggers spelled with a call, a sigil, an acronym or a digit
+
+The whole-word logger rule split a receiver only on `_`, `-` and camelCase, so `logging.getLogger(__name__).error(…)`,
+`get_logger()`, `structlog.get_logger()`, `@logger`, `$logger`, `this.#logger`, `HTTPLogger`, `logger2`, `mylog` and
+`vlog` were no longer loggers, and Python's gating log-only rows were lost. Trailing call arguments now come off before
+the last `.` segment, words split on every non-alphanumeric byte, at acronym-to-word and at letter-to-digit steps, and a
+logger package spelled as one word (structlog, logfire, logbook, loguru) or log/logger/logging behind a one- or
+two-letter prefix counts (so clog, flog and clogging do too). Longer English words that end in "log" (catalog,
+dialog, backlog, analog, changelog), blog and
+technology still log nothing. Still a miss: a logger behind a longer one-word prefix (`auditlog`).
+
 ### Fixed — review round on train 22: false gates, a stack overflow, silent cuts
 
 - `--quality-delta` log-only handlers: a receiver is a logger only when the first or last word of its last segment is `log`, `logger` or `logging` (words split on `_`, `-` and camelCase), or a known logger package, not whenever it contains "log". A Python `except Exception: store.catalog.write(x)` no longer gates as log-only.

README.md

@@ -12,7 +12,7 @@
 **The ripgrep of AI context. A map before your agent reads the repo — and a check on what it writes.**
 
 Ranked, deterministic call graph: what to touch, what it breaks, which tests to run. On the edit: blast
-radius, tests that reach it, ten quality kinds reporting only what got worse, forgotten co-changes, fields
+radius, tests that reach it, eleven quality kinds reporting only what got worse, forgotten co-changes, fields
 read and written, names that resolve more than one way.
 
 **Just want to use it?** Install it with the one line below, then start each coding session by telling your agent to

docs/LIMITS.md

@@ -37,8 +37,8 @@ None of them truncates anything, so none can be judged by `shown=`/`total=` and
 a disclosure — labelling them OUTPUT would ask for a `capped="1"` that could never honestly fire.
 The distinction was named in review on #108 and the rows below now carry it.
 
-The `class` column below carries that answer where it is known. **127 of 230 caps are classified
-(42 INDEXING, 41 OUTPUT, 44 BOUNDARY); the remaining 103 render `—`, which means NOT YET
+The `class` column below carries that answer where it is known. **128 of 230 caps are classified
+(43 INDEXING, 41 OUTPUT, 44 BOUNDARY); the remaining 102 render `—`, which means NOT YET
 CLASSIFIED — never "neither".** Classifications live in `docs/limits_classes.tsv`, a sidecar with
 a known expiry:
 the tag belongs on the declaration itself, and this file exists only because the round that
@@ -360,7 +360,7 @@ Discloses: **none**
 
 | constant | value | class | note |
 | --- | --- | --- | --- |
-| `kMentionsNameMaxDepth` | `512` | — | — |
+| `kMentionsNameMaxDepth` | `512` | INDEXING | undisclosed: a deeper subtree answers "mentions it", so a log-only handler that deep is never flagged |
 
 ### `src/handoff.h`
 

docs/LINEAGE.md

@@ -81,7 +81,7 @@ left to look like a paper.
 | ONTO — [arXiv:2604.17512](https://arxiv.org/abs/2604.17512) | Genuinely tabular rows re-encode far cheaper as parallel arrays than as repeated per-row markup. | `--format=columnar` (`src/columnar.h`) |
 | Controlled serialization study — [arXiv:2603.03306](https://arxiv.org/abs/2603.03306) | The same compact re-encoding collapses on *nested* data, and the headline savings are measured against pretty-printed JSON — a baseline minified XML already beats. | `--format=columnar` refuses every non-tabular verb with exit 1 instead of degrading quietly; the nested map is never re-encoded |
 | Tokenizer and formatting cost — [arXiv:2508.13666](https://arxiv.org/abs/2508.13666) | A single characters-per-token divisor is 20–35% wrong on code, and the spread is dominated by language, not by the tokenizer. | The per-language calibration behind `est_tokens` — and the decision *not* to vendor a byte-pair table, which would buy exactness for the wrong tokenizer |
-| AI-generated code smells — [arXiv:2605.02741](https://arxiv.org/abs/2605.02741) | Bloat tracks architectural decay almost perfectly, and the named failure modes are god-class growth, inline re-implementation of existing interfaces, and file separation mistaken for cohesion. | Three of `--quality-delta`'s ten kinds: verbosity, duplication, and reuse-decline (`src/quality.h`) |
+| AI-generated code smells — [arXiv:2605.02741](https://arxiv.org/abs/2605.02741) | Bloat tracks architectural decay almost perfectly, and the named failure modes are god-class growth, inline re-implementation of existing interfaces, and file separation mistaken for cohesion. | Three of `--quality-delta`'s eleven kinds: verbosity, duplication, and new-clone-of-reused-helper (`src/quality.h`) |
 | Agent-code maintainability — [arXiv:2606.21804](https://arxiv.org/abs/2606.21804) | Classical complexity did not predict real agent-code maintenance failures. Contract drift and code growth did. | The api-surface kind in `--quality-delta`, and `--edit-check` as a standalone contract check (`src/editcheck.h`) |
 | Metric feedback into the loop — [arXiv:2505.23953](https://arxiv.org/abs/2505.23953) | Feeding measured complexity back after a failed attempt raised Pass@1 from 12.5% to 35.7%; the oracle has to be re-readable, not just a verdict. | `--quality-delta` reports only what a change made *worse*, itemised, in a form meant to go back into the next prompt |
 | ARISE — [arXiv:2605.03117](https://arxiv.org/abs/2605.03117) | Statement-level definition-use edges exposed as a queryable agent primitive lifted Function Recall@1 by 17 points over an unmodified SWE-agent — the graph's granularity floor, not its ranking, was the binding constraint. | `--slice=SYM[:VAR]` and the rung-2 `--slice-flow` (`src/slice.h`) — per-line def/use rows inside one resolved definition, plus the bounded transitive BFS over reaching-definition edges. **Bounded, and the output says so**: name-based, intra-procedural, no alias analysis, no control dependence, disclosed depth truncation; the paper's three-tier graph API remains a design note. The reference implementation is now published ([FARD-Lab/ARISE](https://github.com/FARD-Lab/ARISE), MIT) and was audited line-by-line against this slicer 2026-08-31 — the core reaching-definition rule matches; the divergences (their traversal is seed-variable-filtered and unbounded, ours chains operand variables under a disclosed bound) are recorded in [`EVALS.md`](EVALS.md) |