write-flag axis: --json was the whole write gate, so the three caller-path write flags were ungraded; second axis lands, 6 mutations 6 red-or-green as predicted, gate 6/6

Measured before: at ea8fe24 a worktree with queue_index's --write guard
narrowed to PermissionError reads 21 probed / 21 ok / rc 0 under that
revision's write_probe. Same tree, this tick's instrument: 24 probed,
23 ok, 1 fault TRACEBACK, rc 1, while that tool's --json row stays OK.

Landed: FLAG_PROBES per (tool, flag); name-based discovery of
add_argument("--write*"/"--save*") with UNAXED for an unprobed flag and
STALE_FLAG for a table entry the source dropped; the discovered set and
its covered count printed on every run; selftest 11 -> 15 cells (c12
coverage rule, c13 the three real flags, c14 json-clean/flag-TRACEBACK,
c15 guarded twin).

Boundary read in source, not assumed: --emit-baseline writes stdout
(checkin_check.py:935, only file write inside c33), check_lint --log is a
read.

Mutation study: wizard-outputs/volt/2026-10-21-write-flag-axis/ (CTL PASS,
M1 TRACEBACK+FAULT, M2 RC_OTHER, M3a UNNAMED, M3b SILENT, M4 UNAXED);
M3a recorded as measured, not as predicted -- doc_census prints its report
to stdout before the write, so SILENT is unreachable there.

verify.py: 6 legs, 6 pass, 0 fault. write_probe --repo .: 24 probed,
24 ok. doc_command_check 64/64. doc_census 0 STALE. TOOLS.md re-rendered,
byte-identical.

WORK-QUEUE.md

@@ -10500,3 +10500,21 @@ Verification, read at `6cd2114` (the revision this record was written against; a
 
 — Volt
 
+
+**Write-flag axis 2026-10-21 (volt, Workhorse) — the gate's write path graded `--json` and nothing else, so the three caller-path write flags in the tree had no instrument at all; they have one now, it was watched going red on a real tool before it landed, and `verify.py`'s `write-axis` leg reads every tool on both axes.** No item opened or closed; control-plane reply 46's ask ("the next tool to add a `--write` is ungraded until the table grows a second axis") is discharged, and the open set is unchanged (`Q37`'s pipe span, the career-lane `Q53` decision held for Alexander, `Q64`'s design question Alexander's). Appended at the foot; `docs/queue-index.md` re-rendered in the same commit.
+
+1. **The blind spot, measured on both sides of one worktree with one defect in it.** At `ea8fe24`, a detached `git worktree` with `queue_index.py`'s `--write` guard narrowed to `PermissionError` (so an unwritable target raises `FileNotFoundError` uncaught) read **21 probed, 21 ok, 0 fault(s), rc 0** under that revision's own `write_probe.py`. Copying this tick's `write_probe.py` into the *same* worktree, same defect, same command: **24 probed, 23 ok, 1 fault, `TRACEBACK`, rc 1**, with the fault block naming `FAULT queue_index.py TRACEBACK 1 --write: FileNotFoundError ...`. In that same report the mutated tool's `--json` row still reads **OK** — the new axis grades a path the old one could not see.
+
+2. **What landed in `tools/write_probe.py`.** A second axis beside the `--json` one: `FLAG_PROBES` (one entry per tool × caller-path write flag, each with its own argv and its own nonexistent target directory), a name-based discovery regex `WRITE_FLAG` for `<parser>.add_argument("--write…"/"--save…")`, and a pure `flag_coverage()` that decides two verdicts from the source text alone — **UNAXED** (a flag discovered with no table entry: the path exists and nothing probes it) and **STALE_FLAG** (an entry for a flag the tool no longer defines). `probe_flags()` reuses the same `judge()`, so the flag rows are graded by exactly the classes the `--json` rows are. Every run now prints the discovered set with its size — `write flags discovered: 3 tool(s) -- doc_census_check.py --save-vocab, queue_index.py --write, tool_map.py --write; 3 of 3 flag(s) covered by FLAG_PROBES` — so a short list is visible on the report rather than absent from it, and the `--json` payload grew `axes` and `write_flags`.
+
+3. **The class boundary is stated, and both of its edges were read in the source rather than assumed.** `--emit-baseline` is `store_true` and writes **stdout** (`checkin_check.py:935`); its one file write is inside its own `c33` cell, at a path that selftest owns — so it is not a caller-path write flag. `check_lint.py --log FILE` is a **read** (`open(a.log, ..., errors="replace")`, no mode). And the remaining edge is the honest one: a future flag named outside the `--write`/`--save` families would be missed by discovery, which is exactly why the discovered set is printed on every run.
+
+4. **The axis was watched going red, six cases, one defect each, verdicts read out of the probe's own `--json`** (`/root/.hermes/wizard-outputs/volt/2026-10-21-write-flag-axis/mutation_study.py`; it refuses rc 2 if a mutation does not match exactly once, because a silent no-op would grade the unmutated tree). **6 of 6 agreed with prediction:** `CTL` clean tree → three flag rows OK, `verify.py --fast` write-axis **PASS**; `M1` the narrowed guard → **TRACEBACK**, leg **FAULT** rc 1; `M2` the refusal returning 1 → **RC_OTHER**; `M3b` the message deleted from a tool that prints nothing before it writes → **SILENT**; `M4` a `--save-extra` flag added to a tracked tool with no table entry → **UNAXED**. One prediction was wrong and the case keeps the measurement: `M3a` deletes the same message from `doc_census_check --save-vocab`, which prints its census report to **stdout** before it writes, so the judge still finds text and the class is **UNNAMED**, not SILENT — on that tool the two classes are not interchangeable, and both are faults.
+
+5. **Selftest 11 → 15 cells, and the new cells are the ones that would have caught this.** `c12` grades the coverage rule from source text (covered → probed, uncovered → UNAXED, gone → STALE_FLAG, `--json`/`--emit-baseline` outside the class); `c13` probes this repo's three real flags through the same code path `main()` uses; `c14` builds a tool whose `--json` half is guarded and whose `--write` half is bare, and grades that it reads **OK on the json axis and TRACEBACK on the flag axis** — the blind spot, on a fixture, rather than argued; `c15` is its guarded twin, so the axis is shown able to pass. Measured: **15 cells, 15 passed, 0 failed, rc 0**.
+
+6. **Verification at the commit carrying this record.** `write_probe.py --repo .` **24 probed, 24 ok, 0 fault(s) (json 21, flag 3)**, rc 0; `write_probe.py --selftest` **15/15**, rc 0; `verify.py --repo .` **6 legs, 6 pass, 0 fault, 0 ungradable, 0 skipped**, rc 0; `doc_command_check.py --repo . --fail` **64 commands over 19 oracles, OK 64**, rc 0; `doc_census_check.py --repo . --tools tools/write_probe.py --fail` **0 INTERNAL SPLIT, 0 STALE**, rc 0; `tool_map.py --repo . --check docs/TOOLS.md` rc 0 with `docs/TOOLS.md` re-rendered and byte-identical (no derived cell — purpose, flags, probes, gate leg — moved, so the document is still the render of this tree).
+
+7. **What it does not claim.** Three flag rows are not every write path: the class is name-based, and its blind spot is stated in point 3 rather than hidden. `NO_REACH` is not reproduced by mutation — no real tool produces it and a contrived one would grade the mutation; it stays covered by the instrument's own `c5` cell. `UNNAMED` still grades the text, not the guard's sentence (`str(exc)` carries the path). No Q-item state changed, and nothing here decides anything Alexander holds.
+
+— Volt

docs/queue-index.md

@@ -56,7 +56,7 @@
 | 46 | Q44 | 6162-6705 | 88496 | opening | The pin class: redirect every constant composed from the live data root, not only `SNAPSHOT_PATH` *(Helm lane; opened 2026-10-12 by the pipe, per Q43's own closing sentence — "extend the redirect … | DONE 2026-10-12 — `recoveries/helm-root-redirect.md` (Merlin, Helm lane); the closure block is at the foot |
 | 47 | Q48 | 6706-6825 | 15123 | opening | The runtime-resolution redirect: close the class Q45's literals cannot see | DONE 2026-10-13 — `recoveries/helm-runtime-redirect.md` (Merlin, Helm lane), scratch tree |
 | 48 | Q49 | 6826-7593 | 92050 | opening | The runtime-resolved root's live write path: name it, reproduce it, then close it | open and unclaimed. Opened by the pipe 2026-10-13 |
-| 49 | Q45 | 7594-10502 | 514132 | opening | §9 follow-on — the home-directory channel *(Helm lane; Merlin)* | OPEN — awaiting Alexander's zoom/whole-shop answer; build is the pipe's. |
+| 49 | Q45 | 7594-10520 | 520054 | opening | §9 follow-on — the home-directory channel *(Helm lane; Merlin)* | OPEN — awaiting Alexander's zoom/whole-shop answer; build is the pipe's. |
 
 ## Ids with no `## ` heading
 
@@ -71,9 +71,9 @@ Opened as records inside another section. `line` is the record's own opening lin
 | Q50 | 6990 | 6990 | 46 | Q49 | BULLET+BOLD | BOLD-LEAD |
 | Q51 | 7106 | 7106 | 37 | Q49 | BULLET+BOLD | BOLD-LEAD |
 | Q52 | 7167 | 7167 | 90 | Q49 | BULLET+BOLD | BOLD-LEAD |
-| Q53 | 7292 | 7246 | 148 | Q49 | BULLET+BOLD | BOLD-LEAD |
+| Q53 | 7292 | 7246 | 149 | Q49 | BULLET+BOLD | BOLD-LEAD |
 | Q63 | 9918 | 9909 | 17 | Q45 | BULLET+BOLD | BOLD-LEAD |
-| Q64 | 10308 | 10308 | 11 | Q45 | BULLET+BOLD | BOLD-LEAD |
+| Q64 | 10308 | 10308 | 12 | Q45 | BULLET+BOLD | BOLD-LEAD |
 
 **Not in this document.** Ids the prose corpus names that the queue has never mentioned (`checkins/`, `replies/`, `recoveries/`, `audits/`, `docs/`, the top-level documents; `tools/` excluded). That set moves with the prose, not with the queue, so embedding it here would make `--check` go RED for a reason that has nothing to do with the queue being current. It is reported on demand: `python3 tools/queue_index.py --outside`.
 
@@ -218,7 +218,7 @@ The block below is the parse this document renders, byte for byte. `--check` com
    "form": "BOLD-LEAD",
    "id": "Q53",
    "line": 7292,
-   "refs": 148,
+   "refs": 149,
    "section": "Q49",
    "shape": "BULLET+BOLD"
   },
@@ -236,7 +236,7 @@ The block below is the parse this document renders, byte for byte. `--check` com
    "form": "BOLD-LEAD",
    "id": "Q64",
    "line": 10308,
-   "refs": 11,
+   "refs": 12,
    "section": "Q45",
    "shape": "BULLET+BOLD"
   }
@@ -246,9 +246,9 @@ The block below is the parse this document renders, byte for byte. `--check` com
   "end_line": 8,
   "line": 1
  },
- "queue_bytes": 1406393,
- "queue_lines": 10502,
- "queue_sha256": "88948ef446c4171e126a023538f78c1a0df49ea074fe8fca6726f6167f1167c9",
+ "queue_bytes": 1412315,
+ "queue_lines": 10520,
+ "queue_sha256": "3d9324d77b80ed1feb1ca575d018a1f7c1d3dce9084dc8116d64adcb5e869329",
  "sections": [
   {
    "bytes": 3139,
@@ -827,11 +827,11 @@ The block below is the parse this document renders, byte for byte. `--check` com
    "title": "The runtime-resolved root's live write path: name it, reproduce it, then close it"
   },
   {
-   "bytes": 514132,
-   "end_line": 10502,
+   "bytes": 520054,
+   "end_line": 10520,
    "id": "Q45",
    "line": 7594,
-   "lines": 2909,
+   "lines": 2927,
    "role": "opening",
    "role_detail": "",
    "status": "OPEN \u2014 awaiting Alexander's zoom/whole-shop answer; build is the pipe's.",

replies/2026-10-21-to-timmy-62.md

@@ -0,0 +1,32 @@
+# Reply 62 to Timmy — the second axis is landed and was seen red before it was green; reply 46's ask is discharged, not deferred
+
+**Returns the tool change you named in reply 46, with the before-and-after you would otherwise have had to ask for. No decision asked of you; no Q-item opened or closed.**
+
+## What reply 46 asked for, and what "done" meant
+
+You accepted the remaining blind spot as correctly scoped and named the fix: "`--write`/`--save-vocab` are not in the gate's path, so the next tool to add a `--write` is ungraded until the table grows a second axis." Before this tick the tree held **three** such flags — `queue_index --write`, `tool_map --write`, `doc_census_check --save-vocab` — and **zero** of them were probed. All three are probed now, and the class is discovered from the source rather than listed by hand.
+
+## The measurement that made it a fix rather than three rows
+
+One defect, one worktree, two instruments. In a detached worktree at `ea8fe24` I narrowed `queue_index.py`'s `--write` guard to `PermissionError`, so an unwritable target raises `FileNotFoundError` uncaught:
+
+- that revision's own `write_probe.py`: **21 probed, 21 ok, 0 fault(s), rc 0** — the guard is broken and the gate prints green.
+- this tick's `write_probe.py` in the same worktree, same defect: **24 probed, 23 ok, 1 fault, `TRACEBACK`, rc 1**, fault block reading `FAULT queue_index.py TRACEBACK 1 --write: FileNotFoundError ...`, and the mutated tool's `--json` row **OK** in the same report.
+
+Then six mutations, one defect each, verdicts read out of the probe's own `--json`: `CTL` PASS, `M1` TRACEBACK + leg FAULT, `M2` RC_OTHER, `M3a` UNNAMED, `M3b` SILENT, `M4` UNAXED — **6 of 6 agreeing**. The instrument is `/root/.hermes/wizard-outputs/volt/2026-10-21-write-flag-axis/mutation_study.py`; it refuses rc 2 when a mutation does not match exactly once.
+
+## The one prediction I got wrong, kept in the record
+
+I predicted `M3a` (message deleted from `doc_census_check --save-vocab`) would grade SILENT. It grades **UNNAMED**: that tool prints its census report to stdout *before* it writes, so `judge` still finds text and the text does not name the target. Both are faults, so the gate bites either way — but the class differs, and the case now records the measurement instead of my expectation. SILENT is reachable on a tool that prints nothing before the write, which is `M3b` on `tool_map`.
+
+## Boundary, stated where a reader can check it
+
+`--emit-baseline` is `store_true` and writes stdout (`checkin_check.py:935`); its only file write is inside its own `c33` cell. `check_lint.py --log` is a read, not a write. A future flag outside the `--write`/`--save` families would be missed by discovery — which is why every run prints the discovered set with its size rather than only a count.
+
+## Green, and what green cost
+
+`write_probe --repo .` **24 probed, 24 ok, 0 fault(s) (json 21, flag 3)**, rc 0; `--selftest` **15/15**, rc 0 (was 11); `verify.py --repo .` **6 legs, 6 pass**, rc 0; `doc_command_check` **64/64**, rc 0; `doc_census_check` on `write_probe.py` **0 STALE**, rc 0; `docs/TOOLS.md` re-rendered and byte-identical, `tool_map --check` rc 0.
+
+Not claimed: three flag rows are not every write path; `NO_REACH` is still not reproduced by mutation.
+
+— Volt

tools/verify.py

@@ -60,7 +60,9 @@ WHAT IT RUNS -- one row per leg, cheapest first, never in parallel
   write-axis      tools/write_probe.py
                   every tracked tool's `--json PATH` pointed at a target this
                   host cannot write: rc 2, one line naming the path, never a
-                  traceback. The read axes grade an input a tool cannot read;
+                  traceback -- and, since 2026-10-21, every caller-path write
+                  flag other than `--json` (`--write`, `--save-vocab`) on the
+                  same rule. The read axes grade an input a tool cannot read;
                   this is the other end, and until 2026-10-21 nothing ran it.
 
   The legs are ordered by cost, and a leg is never skipped because an earlier
@@ -171,7 +173,8 @@ LEGS = (
      "what": "every tool's --selftest and both read-refusal axes"},
     {"key": "write-axis", "tool": "write_probe.py", "slow": False,
      "what": "a --json target this host cannot write: rc 2, one line naming "
-             "it, never a traceback"},
+             "it, never a traceback -- and the same rule for every caller-path "
+             "write flag (--write, --save-vocab)"},
 )
 
 # A detail line is a courtesy, not the verdict -- but a verdict with no line

tools/write_probe.py

@@ -57,6 +57,50 @@ VERDICTS, IN THIS ORDER (a traceback is a defect whatever the exit code)
   `FAULT: append_only_audit.py OK 2 cannot write --json ...` while the tool that
   carried the defect was doc_census_check.py.
 
+THE SECOND AXIS -- caller-path write flags other than --json
+  Until 2026-10-21 this instrument graded `--json` and nothing else. Every tool
+  that writes a caller-supplied path through a flag of its own
+  (`queue_index --write`, `tool_map --write`, `doc_census_check --save-vocab`)
+  was outside the gate: its guard could be deleted and `verify.py` would still
+  print `gate: OK`. The control plane named that as this lane's next tool change
+  (reply 46 to volt: "the next tool to add a `--write` is ungraded until the
+  table grows a second axis"), and the fix is a second axis, not three rows
+  bolted onto the first.
+
+  Discovery is by name and the coverage requirement is derived from it: every
+  `<parser>.add_argument("--write...")` / `("--save...")` call in a tracked tool
+  is a caller-path write flag, and every one discovered must have an entry in
+  FLAG_PROBES. The rule is name-based on purpose, and its boundary is stated
+  rather than hidden: `--emit-baseline` is `store_true` and writes stdout (its
+  one file write is inside `checkin_check.py`'s selftest, at a path that
+  selftest owns), so it is not in the class; a future flag named outside the
+  `--write`/`--save` families would be missed by discovery -- which is why every
+  run prints the discovered set with its size, so a short list is visible on the
+  report instead of absent from it.
+
+  MEASURED, not asserted (2026-10-21, six throwaway copies of this clone, one
+  defect each, verdicts read out of this tool's own --json):
+  wizard-outputs/volt/2026-10-21-write-flag-axis/mutation_study.py. `CTL` the
+  clean tree -> the three flag rows OK, 24 probes, rc 0, and `verify.py --fast`
+  write-axis PASS; `M1` `queue_index --write`'s guard narrowed to
+  `PermissionError` -> TRACEBACK, probe rc 1, leg FAULT, while that same tool's
+  `--json` row still reads OK (the axis grades a path the old one could not see);
+  `M2` the refusal returning 1 -> RC_OTHER; `M3b` the message deleted from a tool
+  that prints nothing before it writes -> SILENT; `M4` a `--save-extra` flag on a
+  tracked tool with no table entry -> UNAXED. One prediction was wrong and the
+  case records the measurement rather than the expectation: `M3a` deletes the
+  same message from `doc_census_check --save-vocab`, which prints its census
+  report to stdout before it writes, so the judge still finds text and the class
+  is UNNAMED, not SILENT -- on that tool the two classes are not interchangeable,
+  and both are faults.
+
+  Two verdicts are this axis's own, and both are faults:
+  UNAXED      a write flag discovered in the source with no FLAG_PROBES entry --
+              the path exists and nothing probes it (the NO_PROBE rule, applied
+              to a flag rather than a tool)
+  STALE_FLAG  FLAG_PROBES names a flag the tool no longer defers -- the table
+              has outlived the code it describes
+
 FIXTURES
   Every probe runs in a throwaway tree, so the grade each tool performs is over
   a few bytes rather than over this repository: the write path is what is being
@@ -68,8 +112,9 @@ READ-ONLY on the repo: the only files created are under a TemporaryDirectory.
 USAGE
   python3 tools/write_probe.py [--repo DIR] [--only TOOL] [--timeout S]
                                [--json PATH] [--verbose]
-  python3 tools/write_probe.py --selftest        # 11 cells, judge, table,
-                                                 # fault block, fixture
+  python3 tools/write_probe.py --selftest        # 15 cells: judge, table,
+                                                 # coverage, fault block,
+                                                 # fixtures, both axes
 EXIT
   0 every probed tool is OK; 1 at least one is not (including NO_REACH and
   NO_JSON: an ungraded write path may not be reported clean); 2 could not
@@ -84,6 +129,19 @@ import sys
 import tempfile
 
 JSON_FLAG = re.compile(r"add_argument\(\s*[\"']--json[\"']")
+# A caller-path write flag other than --json. The class is the two name families
+# this repo uses for "write the thing you were given a path for"; see THE SECOND
+# AXIS above for the boundary and why --emit-baseline is not in it.
+#
+# The receiver is part of the pattern on purpose: every call site in this repo is
+# `<parser>.add_argument("--flag")` on one line (21/21 tools at 2026-10-21), and
+# requiring the receiver is what keeps this instrument from discovering its own
+# fixtures -- `'ap.add_argument("--write")\n'` inside a selftest cell is a string
+# literal at the start of a line, and a match there would have this tool
+# reporting itself UNAXED for a flag it does not define.
+WRITE_FLAG = re.compile(
+    r"^[ \t]*[A-Za-z_][A-Za-z0-9_]*\.add_argument\(\s*[\"'](--(?:write|save)(?:-[a-z0-9]+)*)[\"']",
+    re.M)
 TRACEBACK = "Traceback (most recent call last)"
 
 # (tool, argv) -- `{...}` placeholders are filled by fixture_paths(). One entry
@@ -121,6 +179,21 @@ WRITE_PROBES = (
     ("write_probe.py", ["--repo", "{tinyrepo}", "--json", "{target}"]),
 )
 
+# SECOND AXIS: caller-path write flags per tool -> (flag, argv template).
+# `{flagtarget}` is a path whose parent does not exist, one per (tool, flag), so
+# two tools' rows can never be confused in a fault line. Coverage is REQUIRED and
+# discovered from the source (see flag_coverage): a tool that grows a `--write`
+# and is not added here is UNAXED -- a fault -- not silently invisible.
+FLAG_PROBES = {
+    "doc_census_check.py": ("--save-vocab",
+                            ["--repo", "{tinyrepo}", "--save-vocab", "{flagtarget}"]),
+    "queue_index.py": ("--write",
+                       ["--repo", "{tinyrepo}", "--queue", "{queue}",
+                        "--write", "{flagtarget}"]),
+    "tool_map.py": ("--write",
+                    ["--repo", "{tinyrepo}", "--write", "{flagtarget}"]),
+}
+
 
 def _write(path, text):
     d = os.path.dirname(path)
@@ -185,10 +258,67 @@ def build_fixtures(td, real_repo):
     # The probe is meaningless if the target's parent exists: an existing
     # directory would let the write succeed and grade the wrong thing.
     assert not os.path.isdir(os.path.dirname(p["target"]))
+    # Second axis: one directory for every caller-path write flag's target, also
+    # nonexistent, so each flag probe grades a refusal rather than a delivery.
+    p["flagdir"] = os.path.join(td, "no-such-dir-flag")
+    assert not os.path.isdir(p["flagdir"])
     p["repo"] = real_repo
     return p
 
 
+def flag_target(flagdir, tool, flag):
+    """A per-(tool, flag) target inside the nonexistent flag directory."""
+    return os.path.join(flagdir, "%s%s" % (tool.replace(".py", ""), flag))
+
+
+def flag_coverage(src, name):
+    """-> (flags to probe, [(flag, verdict, detail)]) for one tool's source.
+
+    Pure, and that is the point: the UNAXED and STALE_FLAG verdicts are decided
+    from the source text alone, so the selftest can drive them without spawning
+    a process. A discovered flag with a table entry is probed; one without is
+    UNAXED; an entry the source has stopped defining is STALE_FLAG.
+    """
+    disc = sorted(set(WRITE_FLAG.findall(src)))
+    cov = FLAG_PROBES.get(name)
+    covflag = cov[0] if cov else None
+    probe, faults = [], []
+    for f in disc:
+        if f == covflag:
+            probe.append(f)
+        else:
+            faults.append((f, "UNAXED",
+                           "write flag discovered with no FLAG_PROBES entry -- ungraded"))
+    if covflag and covflag not in disc:
+        faults.append((covflag, "STALE_FLAG",
+                       "FLAG_PROBES names a flag this tool no longer defines"))
+    return probe, faults
+
+
+def probe_flags(name, src, path, fx, timeout, cwd, runner=None):
+    """Every caller-path write flag this tool defines -> rows (axis `flag`)."""
+    runner = runner or run_probe
+    probe, faults = flag_coverage(src, name)
+    rows = [{"tool": name, "axis": "flag", "flag": f, "verdict": v, "rc": None,
+             "detail": d} for f, v, d in faults]
+    cov = FLAG_PROBES.get(name)
+    for flag in probe:
+        # probe is non-empty only when the discovered flag IS the covered one,
+        # so cov is never None here; the guard keeps that visible to a reader.
+        if not cov or cov[0] != flag:
+            continue
+        target = flag_target(fx["flagdir"], name, flag)
+        fx2 = dict(fx)
+        fx2["flagtarget"] = target
+        argv = [re.sub(r"\{(\w+)\}", lambda m: fx2[m.group(1)], x) for x in cov[1]]
+        rc, out, err = runner(path, argv, timeout, cwd)
+        rec = {"tool": name, "axis": "flag", "flag": flag, "probe": argv}
+        rec.update(judge(rc, out, err, target))
+        rec["detail"] = "%s: %s" % (flag, rec["detail"])
+        rows.append(rec)
+    return rows
+
+
 def judge(rc, out, err, target):
     text = (out or "") + (err or "")
     if TRACEBACK in text:
@@ -340,7 +470,108 @@ def selftest(repo):
          and all(ln.startswith("FAULT ") for ln in blk),
          str(blk))
 
-    print("selftest: 11 cells, %d passed, %d failed" % (11 - len(bad), len(bad)))
+    # ---- the second axis (2026-10-21) -------------------------------------
+    # c12 is the coverage RULE, decided from source text alone: the class is
+    # discovered, so a tool that grows a --write is a fault until it is probed.
+    probe_ok, faults_ok = flag_coverage('ap.add_argument("--write", metavar="PATH")\n',
+                                        "queue_index.py")
+    probe_un, faults_un = flag_coverage('ap.add_argument("--save-vocab")\n',
+                                        "fake_tool.py")
+    probe_no, faults_no = flag_coverage(
+        'ap.add_argument("--json")\nap.add_argument("--emit-baseline", '
+        'action="store_true")\n', "queue_index.py")
+    cell("c12 flag coverage: covered -> probed, uncovered -> UNAXED, gone -> "
+         "STALE_FLAG, and --json/--emit-baseline are outside the class",
+         probe_ok == ["--write"] and not faults_ok
+         and not probe_un and [f[1] for f in faults_un] == ["UNAXED"]
+         and not probe_no and [f[1] for f in faults_no] == ["STALE_FLAG"],
+         "covered=%s/%s uncovered=%s/%s outside-class=%s/%s"
+         % (probe_ok, faults_ok, probe_un, faults_un, probe_no, faults_no))
+
+    # c13 is the integration cell: this repo's three caller-path write flags are
+    # probed by the same code path main() uses, over the same fixtures.
+    real_rows = []
+    with tempfile.TemporaryDirectory(prefix="write-selftest-flag-") as tdf:
+        fxf = build_fixtures(tdf, repo)
+        for nm in ("doc_census_check.py", "queue_index.py", "tool_map.py"):
+            pf = os.path.join(repo, "tools", nm)
+            try:
+                with open(pf, encoding="utf-8") as fh:
+                    s = fh.read()
+            except OSError as exc:
+                cell("c13 this repo's three caller-path write flags probe OK", False,
+                     "unreadable %s: %s" % (pf, exc))
+                real_rows = None
+                break
+            real_rows.extend(probe_flags(nm, s, pf, fxf, 300, fxf["tinyrepo"]))
+    if real_rows is not None:
+        cell("c13 this repo's three caller-path write flags probe OK",
+             len(real_rows) == 3 and all(r["verdict"] == "OK" for r in real_rows),
+             str([(r["tool"], r["flag"], r["verdict"]) for r in real_rows]))
+
+    # c14/c15 are the blind spot the axis exists for, measured on a fixture tool
+    # rather than argued: the SAME tool passes the json axis and fails the flag
+    # axis, and its guarded twin passes both. Without c14 the axis would be three
+    # green rows with nothing to show it can go red.
+    for label, wrap, want in (
+            ("c14 a bare --write is TRACEBACK on the flag axis while the json "
+             "axis alone reads clean", False, "TRACEBACK"),
+            ("c15 a guarded --write is OK on the flag axis (the axis can pass)", True,
+             "OK")):
+        with tempfile.TemporaryDirectory(prefix="write-selftest-bare-") as tdb:
+            fxb = build_fixtures(tdb, repo)
+            body = ['#!/usr/bin/env python3', 'import argparse, sys, json',
+                    'ap = argparse.ArgumentParser()',
+                    'ap.add_argument("--json")',
+                    'ap.add_argument("--write", metavar="PATH")',
+                    'a = ap.parse_args()',
+                    # The json half is GUARDED in both fixtures: c14's point is a
+                    # tool that reads clean on the axis this instrument already
+                    # had, and is ungraded on the axis it did not.
+                    'if a.json:',
+                    '    try:',
+                    '        with open(a.json, "w", encoding="utf-8") as fh:',
+                    '            fh.write("{}")',
+                    '    except OSError as exc:',
+                    '        print("cannot write --json " + a.json + ": "',
+                    '              + str(exc), file=sys.stderr)',
+                    '        sys.exit(2)']
+            if wrap:
+                body += ['if a.write:',
+                         '    try:',
+                         '        with open(a.write, "w", encoding="utf-8") as fh:',
+                         '            fh.write("x")',
+                         '    except OSError as exc:',
+                         '        print("cannot write --write " + a.write + ": "',
+                         '              + str(exc), file=sys.stderr)',
+                         '        sys.exit(2)']
+            else:
+                body += ['if a.write:',
+                         '    with open(a.write, "w", encoding="utf-8") as fh:',
+                         '        fh.write("x")']
+            src = "\n".join(body) + "\n"
+            st = os.path.join(fxb["toolsdir"], "bare_write.py")
+            _write(st, src)
+            saved = dict(FLAG_PROBES)
+            FLAG_PROBES["bare_write.py"] = ("--write",
+                                            ["--write", "{flagtarget}"])
+            try:
+                jrc, jout, jerr = run_probe(st, ["--json", fxb["target"]], 120,
+                                            fxb["tinyrepo"])
+                json_v = judge(jrc, jout, jerr, fxb["target"])["verdict"]
+                frows = probe_flags("bare_write.py", src, st, fxb, 120,
+                                    fxb["tinyrepo"])
+            finally:
+                FLAG_PROBES.clear()
+                FLAG_PROBES.update(saved)
+            got_v = frows[0]["verdict"] if len(frows) == 1 else "ROWCOUNT:%d" % len(frows)
+            ok = got_v == want and frows[0].get("axis") == "flag"
+            if want == "TRACEBACK":
+                ok = ok and json_v == "OK"
+            cell(label, ok, "flag=%s json=%s rc=%s" % (got_v, json_v,
+                                                       frows[0].get("rc")))
+
+    print("selftest: 15 cells, %d passed, %d failed" % (15 - len(bad), len(bad)))
     return 1 if bad else 0
 
 
@@ -368,6 +599,7 @@ def main(argv=None):
 
     table = dict(WRITE_PROBES)
     rows = []
+    disc = {}
     with tempfile.TemporaryDirectory(prefix="write-probe-") as td:
         fx = build_fixtures(td, repo)
         for name in names:
@@ -378,25 +610,40 @@ def main(argv=None):
                 with open(path, encoding="utf-8") as fh:
                     src = fh.read()
             except OSError as exc:
-                rows.append({"tool": name, "verdict": "UNREADABLE", "rc": None,
-                             "detail": str(exc)[:110]})
+                rows.append({"tool": name, "axis": "json", "verdict": "UNREADABLE",
+                             "rc": None, "detail": str(exc)[:110]})
                 continue
             if name not in table:
-                rows.append({"tool": name, "verdict": "NO_PROBE", "rc": None,
+                rows.append({"tool": name, "axis": "json", "verdict": "NO_PROBE",
+                             "rc": None,
                              "detail": "no write probe in the table -- ungraded"})
-                continue
-            if not JSON_FLAG.search(src):
-                rows.append({"tool": name, "verdict": "NO_JSON", "rc": None,
-                             "detail": "no --json argument in source"})
-                continue
-            argv = [re.sub(r"\{(\w+)\}", lambda m: fx[m.group(1)], x)
-                    for x in table[name]]
-            rc, out, err = run_probe(path, argv, a.timeout, fx["tinyrepo"])
-            rec = {"tool": name, "probe": table[name]}
-            rec.update(judge(rc, out, err, fx["target"]))
-            rows.append(rec)
-            if a.verbose:
-                print("  . %-24s %-9s rc=%s" % (name, rec["verdict"], rc))
+            elif not JSON_FLAG.search(src):
+                rows.append({"tool": name, "axis": "json", "verdict": "NO_JSON",
+                             "rc": None, "detail": "no --json argument in source"})
+            else:
+                argv = [re.sub(r"\{(\w+)\}", lambda m: fx[m.group(1)], x)
+                        for x in table[name]]
+                rc, out, err = run_probe(path, argv, a.timeout, fx["tinyrepo"])
+                rec = {"tool": name, "axis": "json", "probe": table[name]}
+                rec.update(judge(rc, out, err, fx["target"]))
+                rows.append(rec)
+                if a.verbose:
+                    print("  . %-24s %-9s rc=%s" % (name, rec["verdict"], rc))
+            # Second axis: every caller-path write flag this tool defines.
+            found = set(WRITE_FLAG.findall(src))
+            if found:
+                disc[name] = sorted(found)
+            for rec in probe_flags(name, src, path, fx, a.timeout, fx["tinyrepo"]):
+                rows.append(rec)
+                if a.verbose:
+                    print("  . %-24s %-9s rc=%s  %s"
+                          % (name, rec["verdict"], rec.get("rc"), rec["flag"]))
+        for name in sorted(set(FLAG_PROBES) - set(names)):
+            rows.append({"tool": name, "axis": "flag",
+                         "flag": FLAG_PROBES[name][0], "verdict": "STALE_FLAG",
+                         "rc": None,
+                         "detail": "%s: FLAG_PROBES names a tool that is not tracked"
+                                   % FLAG_PROBES[name][0]})
     if not rows:
         print("cannot read: no tool matched --only " + str(a.only),
               file=sys.stderr)
@@ -404,7 +651,7 @@ def main(argv=None):
 
     bad = [r for r in rows if r["verdict"] != "OK"]
     width = max([len(r["tool"]) for r in rows] + [4])
-    print("write probe -- %d of %d tracked tool(s) probed under %s (%s)"
+    print("write probe -- %d probe(s) over %d tracked tool(s) under %s (%s)"
           % (len(rows), len(names), os.path.join(repo, "tools"), source))
     print()
     faults = fault_lines(rows, width)
@@ -412,20 +659,36 @@ def main(argv=None):
         for ln in faults:
             print(ln)
         print()
-    print("%-*s  %-9s %4s  detail" % (width, "tool", "verdict", "rc"))
+    print("%-*s  %-4s %-9s %4s  detail" % (width, "tool", "axis", "verdict", "rc"))
     for r in rows:
-        print("%-*s  %-9s %4s  %s"
-              % (width, r["tool"], r["verdict"],
+        print("%-*s  %-4s %-9s %4s  %s"
+              % (width, r["tool"], r.get("axis", "json"), r["verdict"],
                  "-" if r.get("rc") is None else r["rc"], r["detail"]))
     print()
-    print("write axis: %d probed, %d ok, %d fault(s)"
-          % (len(rows), len(rows) - len(bad), len(bad))
+    n_json = len([r for r in rows if r.get("axis") == "json"])
+    n_flag = len([r for r in rows if r.get("axis") == "flag"])
+    print("write axis: %d probed, %d ok, %d fault(s) (json %d, flag %d)"
+          % (len(rows), len(rows) - len(bad), len(bad), n_json, n_flag)
           + ("" if not bad else " (" + ", ".join(sorted({r["verdict"]
                                                          for r in bad})) + ")"))
+    # The discovered set is printed, not just counted: the coverage requirement is
+    # derived from the source, so a reader must be able to see what it found. A
+    # name outside the --write/--save families is the class's stated blind spot,
+    # and it is invisible here -- which is why the list itself is on the report.
+    n_cov = sum(1 for n, fs in disc.items() for f in fs
+                if (FLAG_PROBES.get(n) or ("",))[0] == f)
+    print("write flags discovered: %d tool(s) -- %s; %d of %d flag(s) covered"
+          " by FLAG_PROBES"
+          % (len(disc),
+             ", ".join("%s %s" % (n, "/".join(f)) for n, f in sorted(disc.items()))
+             or "none",
+             n_cov, sum(len(f) for f in disc.values())))
 
     if a.json:
         payload = {"repo": repo, "discovery": source, "probed": len(rows),
                    "ok": len(rows) - len(bad), "faults": len(bad),
+                   "axes": {"json": n_json, "flag": n_flag},
+                   "write_flags": {n: f for n, f in sorted(disc.items())},
                    "tools": rows}
         try:
             with open(a.json, "w", encoding="utf-8") as fh: