v0.4.0

This section records the user-visible changes since v0.3.0. Internal test, CI
and docs-only commits are omitted unless they change a published surface.

Read the first entry under Changed before you upgrade. One rename in this range
breaks consumers silently. Nothing raises an error and nothing logs a warning.
A program that matched the old value stops matching and reports nothing.

Changed

• The domain glossary is renamed from CONTEXT.md to GLOSSARY.md (#722).
  Pocock v1.3.1 names the domain glossary GLOSSARY.md; a reader following
  any live pointer now lands on that file. Git recognizes the move as a
  rename. The title line names the glossary; the body is byte-identical.
  docs/agents/domain.md is replaced with the v1.3.1
  setup-matt-pocock-skills template, and AGENTS.md names GLOSSARY.md on
  both live instruction lines. Existing CHANGELOG entries that name the old
  path stay as written.

• BREAKING. report_schema_version goes from 1.4.0 to 2.0.0, because the
  aggregation_method value bh_fdr_fallback is renamed to
  bounded_pooling_refused (#360, #441, in 03ee27d). The constant lives at
  src/skill_harness/aggregation/report.py:44. The new value is written at
  src/skill_harness/aggregation/fit.py:759. There is no alias on the write
  path and none on the read path. The old string survives only in comments at
  fit.py:255 and report.py:17.

  What breaks. Any consumer that compares against the literal
  "bh_fdr_fallback" stops matching. The comparison returns false rather than
  raising, so the consumer reports nothing instead of failing. Code that
  branches on the value takes the wrong branch. Code that filters on it returns
  an empty result.

  What to do. Match "bounded_pooling_refused". If you must read both
  generations, branch on report_schema_version rather than accepting either
  string, because the two values report different outcomes. The fit is refused
  and pooled at the admission bound, rather than falling back to BH-FDR. A
  consumer that cannot migrate yet can stay on 0.3.0
  (pip install skill-harness==0.3.0). That release reports
  report_schema_version 1.4.0 and still writes "bh_fdr_fallback".

  docs/PRD.md:757 states the rule this follows. Removals, renames, and type
  changes take a major bump, because they break diff skill consumers.

• The runtime dependency floors rise to the values pyproject.toml declares
  at this PR's head. An environment that pins any package below its floor
  no longer satisfies this package's requirements. Measured from
  git diff v0.3.0 HEAD -- pyproject.toml:

  • anthropic>=1.15.0,<2 (was >=1.2.0; the floor passed through 1.5.0 in
    #548 and 1.11.0 in #727, which re-asserts the post-patch bound for
    GHSA-q5f5-3gjm-7mfm and GHSA-w828-4qhx-vxx3, and landed at 1.15.0 in
    #749). The <2 cap is new (#732). anthropic 2.0.0 is a breaking major
    that no audit has reviewed, so an environment that already resolved
    2.x no longer satisfies this package.
  • openai>=3.31.0,<4 (was >=2.41,<4; the lower bound rose in #604 and
    again in #732. 3.17.0 is the first release that requires the patched
    httpx2>=2.12.0; the floor takes Dependabot's 3.31.0, inside the
    audited 3.x range. The <4 cap is unchanged).
  • click>=8.5.0 (was >=8.1; #606).
  • scipy>=1.18.1 (was >=1.11; #729 is the last of the floor bumps in
    the EB-MoM numerical stack).
  • statsmodels>=0.15.0 (was >=0.14; #605).
  • pydantic>=2.14.1 (was >=2.6; #750).
  • tiktoken>=0.14.0,<1.0 (was >=0.7,<1.0; #751. The <1.0 cap is
    unchanged).

  rich did not move. Dev-extra floors moved with Dependabot and are not
  runtime requirements.

• The ablation runner accepts a whole prompt assembly as an arm, not only a
  clause of one card (#554, #615, in 9d956e8a). RunConfig carries a
  declared arm set serialized into runs.config_json. An assembly may name a
  foreign skill body by path. Duplicate assemblies are refused before any
  spend. Arm samples are written to the append-only evidence.arm_samples
  table (migration 1200), and the A41 reconciler sums them with samples.usd.

• Every external-check refusal reason is persisted (#555, #626, #629,
  in 6b6b74a8). Migration 1100's CHECK constraint listed ten
  UnmeasuredSubReason literals; external_check_missing was never added, and
  INSERT OR IGNORE silently dropped the rows. Migration 1300 rebuilds
  clause_run_outcomes with all eleven members, adds a nullable
  sought_oracle column, and recreates the append-only triggers. Historical
  runs that lost those rows stay lost; new runs record the reason.

• A renamed published skill no longer loses its value_class (#601, #618,
  in d629227e). The registry key closure-mode-at-boundaries still pointed
  at the pre-rename card name, so value_class_for("closure-mode") returned
  None and the skill took the unclassified path instead of
  trap-discipline. The key is now closure-mode, and a staleness test
  fails when any non-retired key stops naming a live skill.

• A clause whose scoring axis reaches no comment or clarity outcome is
  refused before sampling, with a clause-specific reason (#555, #626, in
  6996e68d). The blanket BLOCKER-1 refusal is replaced by an
  external_check_missing sub-reason that names the axis.

• The CLI's rich Console reads COLUMNS at render time, not at import
  (#564, #624, in ddf1373e). A module-level Console() froze the width for
  the life of the process, so terminal resize and
  CliRunner(env={"COLUMNS": "200"}) were inert.

• The published receipt pages stop printing machine tokens as values
  (#583, in 330eca7b). A boolean renders yes or no; a None whose
  schema field names its own null case uses that word (value_class renders
  unclassified); everything else uses the absent class treatment. The
  site also gains page types, a shell with a current-page marker, a 404,
  schema-sourced measurement-table labels, and --base-url so the generator
  stops baking its own host into the output. The landing page sits behind
  --landing (default off) and its one instruction is a command that runs.

• The ablation results table gains a gate-2 (registered floor) column
  (#546, #558, in c2fa804). A clause that previously rendered a green
  PASSED now renders PASSED (scalar only), HARM (registered floor),
  EQUIVALENT (registered floor), or UNRESOLVED (registered floor). The
  verdict words did not change meaning. The report now shows which decision
  produced them. Anyone who parses this table by column position must re-read
  it.

• Tie-bearing ablation clause decisions route through the Gate-2 discordant
  rule (#368, #493, in 0274dd1). A clause whose decision turned on a tie
  can receive a different verdict than it did under 0.3.0.

• paired-gate2 refuses a run that carries undecided epochs (#438, #491,
  in 1a5027a). A hazard entry is counted per simple command rather than per
  tool-call blob. When an arm carries undecided epochs, the command emits a
  HAZARD_UNDECIDED refusal and exits 1. A run that previously produced a
  score now produces a typed refusal instead.

• The PyPI package description is replaced (#497, in 3ba4f69). It now
  carries the repository's GitHub About text.

Added

• skill-harness run pi-paired CONFIG [--execute], a paired subject lane
  for the Pi runtime (src/skill_harness/cli/main.py:3452). The default is a
  dry run that performs the pre-spend gate only. --execute runs the Full and
  Null epochs and spends money. The subcommand requires a pi binary that this
  package does not install. Without it the launcher raises
  PiExecutableNotFoundError
  (src/skill_harness/subject/pi/launcher.py:164). The runtime version is
  measured from the binary that ran and is never assumed. No sized Pi
  measurement has been run, so no Pi skill effect is admissible yet.

• SERS 1.4.0 and SERS 1.5.0. 1.4.0 splits the subject and extractor model
  roles and adds subject_identity.subject_model, required only under the
  1.4.0 conditional (#487, in c398cca). 1.5.0 adds four optional top-level
  keys from the On-Irreducibility additions (#526, #547, in cfb5cec). Both
  are additive for existing receipts. A consumer holding an older copy of the
  schema rejects a new receipt, which the standard treats as intended
  generational non-comparability.

• SERS 1.6.0. Separates the verdict from its present validity (#643, #654,
  in c879633, finished in #655/#660, in 282a5064). A receipt reader must
  handle three new optional top-level objects that are absent on pre-1.6.0
  receipts: verdict_scope (required on a 1.6.0 KEEP, with model_id,
  task_family and delivery_mechanism), currentness (VALIDATED requires
  basis FULL_VALIDATION; CARRIED_FORWARD requires verdict_scope.tested_at
  and renders a carry-forward disclaimer; STALE names a newer metric version
  or hash mismatch), and drift_policy. Receipts from different
  sers_version values are not comparable.

• SERS 1.7.0. Adds the online-FDR ledger facts and the claim-level ladder
  to the receipt (#644, #686, in bd3467c; ladder fields from #647, #671, in
  3f8d4a90). A receipt reader must handle two changes. First, a 1.7.0 KEEP
  carries a required multiplicity object (LORDdep procedure, target FDR
  0.05, anytime_valid_p, wealth before/after, batch order) and a
  verdict_scope.control_world_result that must equal pass; a 1.7.0 CUT
  must not carry multiplicity at all. Second, KEEP receipts gain
  designer, designer_independent, family_replication, model_replication
  and claim_level (KEEP / REPLICATED / ROBUST), with the ladder thresholds
  enforced on write.

• The claim-level ladder (#647, #671, in 3f8d4a90). A family record
  carries five mechanism fields and registration checks. Receipts record who
  designed the claim, whether the designer is independent, how many
  independently designed families replicated, and how many materially distinct
  families a robust claim needs (three, with at least one independent
  designer). Render sentences carry the same-designer qualifier.

• Collection screens record what each run actually delivered (#664, #672,
  in ef69c192; #667, #698, in 2787c13c). The delivered-surface extractor
  marks each subject card visible (with position), truncated (with delivered
  length) or dropped, never guessing an empty listing. The collection screen
  gains a seeded random-half arm: the seed and the drawn card names freeze
  into runs.config_json, and a delivery-integrity flag marks a run whose
  intended card was dropped or truncated from the delivered listing.

• The [geometry] extra (#589, in the rendered-geometry harness).
  pip install "skill-harness[geometry]" pulls playwright>=1.63.0, the
  browser behind src/skill_harness/sitegen/geometry.py. The core package
  never imports playwright; the module raises BrowserNotInstalledError with
  the extra named. Installing the distribution is half the job: the Chromium
  build is fetched separately with python -m playwright install chromium.

• Drift-check rows AC-2, AC-3 and AC-4. AC-3 covers the public vacuity
  claims (#543, #565, in f3f2ba9). AC-4 covers workflow configuration (#571,
  in dc7520e). Rows DC-17 (mirror records, #514), DC-18 (Vale version
  pins, #488) and DC-19 (asset pairs, #592) are also live. The drift check
  reports 22 live contracts at this version.

• Storage migrations 1100_clause_run_outcomes.sql and
  1101_structural_covariates.sql (in e047967), plus 1200 for arm samples
  (#615) and 1300 for the clause_run_outcomes CHECK rebuild (#629). All
  create or rebuild tables additively from a v0.3.0 store's point of view;
  1300 is the one that replaces a table in place.

• The release gate reports its own coverage (#576). Every gate that does
  not run prints a named skip line, and both summary lines state how many of
  the eight gates ran. At 0.3.0 the gate prints 7 of 8 gates ran; skipped G6.
  Off the 0.3 minor line it prints 5 of 8 gates ran; skipped G6, G7, G8. G7
  and G8 previously returned in silence, so RELEASE GATE: PASS covered six of
  eight checks and did not say so.

Fixed

• Ingest finds the skill listing where current Claude Code writes it
  (#675, #687, in a9fb533; #674, in 6505b42). Current Claude Code puts the
  Skill-tool listing in a system message, not in the first user message. The
  extractor and the Stage 1A screens reader searched only the first user
  message, so a delivered card was recorded as no_listing and a look
  crashed after both samples completed. Both now scan messages in order until
  the first assistant message and read the listing from the first message of
  any role that holds Claude Code's header. A listing that first appears at or
  after the first assistant message is still not attributed to the run.

• requirements-ci.txt pins urllib3==2.8.0 for
  CVE-2026-97687/97688/97689 (#701, in e235bc6). This is the CI
  constraints file, not a runtime floor in pyproject.toml; the bump closes
  the advisory on every environment that installs from that pin.

• The SERS schema held the UNMEASURED sub-reason vocabulary twice and one
  copy was untested (#578). $defs/rate_or_refusal is referenced by ten
  measurement fields and was short tier2_uncalibrated and
  length_confounded, so a receipt recording either of the runner's
  pre-sampling refusals could not be written. Both copies are now asserted
  equal to UnmeasuredSubReason.

• ExtractedClause.axis is persisted unnormalised, so a padded axis splits one
  axis into two (#504, in 8d5e0d6).

• unmeasured_reason is never persisted, and its vocabulary is disjoint from
  UnmeasuredSubReason (#503, in 710ef78).

• An absent dump column is a difference, not agreement (#360, in 1f5a73e).

• Two mutants share the id M-R1 and two share M-R2, and nothing asserts
  that a mutant id is unique (#536, in af0788f).

• The lint gate does not cover scripts/, and a syntax error reached a commit
  through the gap (#535, in bef52e1).

• pre-commit run --all-files cannot pass on main, because ruff fails on
  prototypes/, which CI never lints (#483, in fe5b3c1).

• test_mint_path_allowlist filters on the absolute path, so it is vacuous in
  a build worktree (#482, in 7284342).

• Receipt pages render no subject_identity, so the 1.4.0 subject_model is
  invisible to a reader (#490, in eebaad8).

• DC-16 selects the tracked set, not a filesystem walk (#471, #473, in
  b06272a).

• The six UNLANDED mirror rows point at an open ticket and no longer pin its
  number (#521, in 658f769).

• The G7 and G8 GitHub reads authenticate with GITHUB_TOKEN (#496, in
  cf250bd). Unauthenticated reads share a 60-per-hour budget per IP, which
  shared CI runners exhaust.

• The dry-run claim is scoped to the subcommand that spends (#472, in
  8dab4d3).

• The frozen class-2 reference is runnable, its copies are pinned, and the R40
  receipt is frozen (#458, in 6225c66).

• numpy stays out of the runtime manifest (#161, #442, in 52ee261).

• Two tests reported the environment as a code failure. One now skips with the
  privilege named, and the other scans where the build actually runs (#484, in
  db3a688).

• The design snapshot describes the banner main draws, not the window it
  retired (#481, in 466e629), and no longer declares the paper surface that
  #308 retired (#485, in 8a4d064).

• README "Measuring for real" snippet now shows what run ablation --execute
  actually requires (#460). --execute also needs --ratification,
  --task-family, and --estimand; the ratification preflight refuses before
  any model call and before the API-key check, which the prior wording did not
  say.

• CHANGELOG 0.3.0 entry for #424 corrected. The ticket did not stay open on
  the negative control; it closed 2026-09-04 once the control was re-seeded to
  satisfy the #403 ruling's qualified-fixture precondition (#432). The 0.3.0
  section otherwise unchanged.

Internal

The bulk of this range is a rebuild of the empirical-Bayes method-of-moments
peel in src/skill_harness/aggregation, 65 commits that change fit.py by 904
lines. The rebuild carries one externally visible consequence, the rename
recorded under Changed. The work is covered by a frozen pre-registration
(b477983), a superseding v2 frozen on mechanism class 2 (e11e4e2,
397e840), a confirmatory run that returned REJECTED at R=1000 on the
committed root (13f0fbb) and NOT_REJECTED on a fresh root (feb318d), and
mutation receipts re-run against the code that ships (36dcc22, 4348275).

Stage 1A and Stage 2 regime simulators for #621/#685, the v3a Null
qualification screen, comparative collection screens, ADR 0002, and the
On-Irreducibility decline records change research findings and internal
records, not a PyPI-observable surface. Dependabot bumps of pinned dev and CI
actions throughout the range follow the same rule.