Repository navigation
feat(rules): RSR 3.1.1 and 3.2.1 read the chora deed first - #1117
Conversation
Owner ruling 2026-11-14, "Full migration now". Each retired descriptile is translated under the mapping spec for its family, then deleted: - META.a2ml -> (meta): last-updated, scoping-first and the five ADRs (id, title, status, date), per meta-ecosystem-to-repo-deed.adoc. - ECOSYSTEM.a2ml -> (ecosystem): project, ecosystem, position-type, purpose. The seven related-projects entries are not translated: their relationships are outside the closed set the (related) row allows. - AGENTIC.a2ml -> (agentic): version, last-updated, permissions and the five integrity axioms (allow-silent-skip stays #f). - NEUROSYM.a2ml -> (neurosym): version, last-updated and the hypatia scan config (scan-depth deep). - PLAYBOOK.a2ml -> (playbook): version, last-updated and the five operational positions. - STATE.a2ml -> (maturity :level alpha), state-v1-decision option B; the lifecycle claim is the existing (status) clause and the journal is archived by reference, not translated. - CLADE.a2ml was already translated (identity, clade, forges, lineage, status); the legacy file is deleted. Fields no table lists are dropped, not re-homed; the deed comments name each one. The deed lints clean (deed_lint.js, plain and --egress). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
…FEST.deed The machine front door is now 0-AI-MANIFEST.deed, pointer-only (A7 ruling Q15): it names hypatia_chora.deed and README.adoc and carries no facts. The A2ML manifest's sections (identity, context tiers, canonical locations, architecture, invariants, lifecycle) do not match the field table of ai-manifest-decision.adoc, so nothing was translated from it; its architecture summary duplicated .claude/CLAUDE.md. Readers moved with it: - MUST.contractile: the structure invariant now preserves the .deed. - www/.well-known/ai.txt and .github/CONTRIBUTING.md point agents at the deed front door and the chora deed. - hypatia-cli-bash.sh and poc-scanner.sh accept 0-AI-MANIFEST.deed first (the legacy names still count for unmigrated repos) and recommend the deed, not an .a2ml file. The deed lints clean (deed_lint.js). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
Owner ruling 2026-11-14, "Full migration now"; A7 section 11, ruling Q7:
a file is translated where a mapping spec exists and deleted where none
does. None of these has a spec in standards 1-formats/deed/mappings/:
- .machine_readable/anchors/ANCHOR.a2ml (it also has no `authority`, which
the (anchor) clause requires and a translator never invents)
- .machine_readable/bot_directives/{coverage,debt,methodology}.a2ml and
their README (template methodology boilerplate, not per-bot directives)
- .machine_readable/contractiles/{bust,dust,intend,must,trust}/*file.a2ml
- .machine_readable/integrations/*.a2ml (four integration notes)
- .machine_readable/dogfooding.a2ml, testing/blitz.a2ml,
merge-orchestration/LEDGER.a2ml (a journal; archived by reference)
- rules.a2ml
docs/history/A2ML-REMOVED.adoc records every removed file with its size,
sha256 and git blob id, what it held and where its content went, plus the
two .a2ml files that stay and why.
References moved with the files: bust.ncl names Bustfile.deed (A7 8.1);
contractile.just says its sources are retired and must not be regenerated
from; GOVERNANCE.adoc points bot configuration at the (bot-directives)
clause; comments in verdict.ex, findings.ncl, schema/README.adoc,
Verdict.idr and ObservedIsNotSound.idr state the rule-lifecycle threshold
without citing the deleted file; design and audit docs point at the ledger;
docs/integration/a2ml.adoc is marked historical.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
…a2ml RuleLoader is a shared reader of standards' hypatia-rules/*.a2ml, which is still the only form those definitions take (A7 section 10 moves them to A5 rule registry entries in both repositories together), so its .a2ml reading stays. Its fixtures need not be checked-in .a2ml files: the three definitions are now ~S heredocs in the test, written to a temporary directory in setup_all. They are byte-identical to the deleted fixtures (verified by diff), and the ten tests pass unchanged. docs/history/A2ML-REMOVED.adoc gains the three removed fixtures. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
… ANCHOR A7 M5: shared readers read the new input first and keep the A2ML only as a fallback. Hypatia's own descriptiles now live in hypatia_chora.deed, and the oracle scored them as missing. RsrConformance (the RSR oracle): - 3.1.2 STATE accepts a top-level (status) clause (state-v1-decision, option B: the one lifecycle claim is that clause); - 3.1.3 META and 3.1.4 ECOSYSTEM accept (meta) and (ecosystem) (meta-ecosystem-to-repo-deed), as 3.1.5-3.1.7 and 3.1.9 already did for their clauses; - 3.1.8 accepts the (anchor) clause (A7 ruling S1: there is no ANCHOR.deed); - 3.2.1 with no record-dialect .a2ml left passes when the repo deed parses, and still fails when there is neither. The legacy .a2ml carriers still count. Nested clauses never do. HonestCompletion: a repo has state when its chora deed carries a top-level (status) clause or it has a STATE.deed; STATE.a2ml and STATE.scm still count. The finding no longer tells a repo to add an .a2ml file. Hypatia's self-score against the bundled criteria: 3.1.2-3.1.4 and 3.2.1 pass from the deed. 3.1.1 (a descriptiles/ directory) now fails: the criterion names the directory, and hypatia has none left. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
- .claude/CLAUDE.md: the "Machine-Readable Artefacts" section told agents to read six descriptiles/*.a2ml files that no longer exist. It now names 0-AI-MANIFEST.deed, hypatia_chora.deed and its clauses, the rules praxis deed and the removal ledger, says A2ML must not be created, and lists the two .a2ml files that remain and why. - README.adoc "Project status" points at hypatia_chora.deed instead of STATE.a2ml and PLAYBOOK.a2ml. - docs/quickstart/llm-warmup-dev.adoc no longer lists .machine_readable/6a2/ *.a2ml as the metadata home. scripts/check-readme-claims.sh and tests/readme-claims-test.sh pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
The script converted STATE/META/ECOSYSTEM .scm checkpoints into new .a2ml files. A2ML is retired and nothing may create one (standards rulings D99, D269c, D312, D313; A7 Part 1 stops the producers). Nothing in the tree called it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
…ing 2026-11-14) The owner ruled that the <repo>_chora.deed clauses satisfy RSR criteria 3.1.1-3.2.2, with .a2ml only as the A7 new-input-first fallback. The canon (standards 0-canon/rsr/rsr-criteria-v2.a2ml) cannot be revised in place: it is A2ML, and changing a criterion to name a new carrier adds a fact to an .a2ml file. So the oracle changes, and the canon text waits for its deed carrier (A7 section 10: a criteria praxis deed under a mapping spec, written by kyX's front end). RsrConformance: - 3.1.1 (descriptiles-dir): deed first, the chora deed parses and carries a top-level descriptile clause ((status) (meta) (ecosystem) (agentic) (neurosym) (playbook) (anchor) (clade)); then .machine_readable/ descriptiles/, which still counts (the legacy home, and where A7 7.1 puts the ALL-CAPS descriptile deeds). Nested clauses do not count. - 3.2.1 (a2ml-valid): every carrier must parse, the deed first. A broken chora deed now fails it even when the remaining record-dialect .a2ml parse (before, it was ignored whenever an .a2ml existed); a valid deed still does not excuse a malformed .a2ml. - 3.1.2-3.1.9: the deed clause is read first and the .a2ml is the fallback (same verdicts; the order now matches A7 M5). Tests: 3.1.1 for each of the eight clauses, the nested / clause-less / broken-deed failures and the directory fallback; 3.2.1 with a broken deed beside a valid .a2ml. The self-application smoke now asserts 3.1.1 and 3.2.1 pass for hypatia. The test deeds' ids move from #u5 (an error in an identity field, owner ruling Q2) to #u8 / #uuid. Hypatia self-score (bundled criteria): 84.08% -> 86.91%, bronze; against the standards canon (2.1.0-draft): 82.71% -> 85.78%. Still failing: 1.2.4, 2.1.4, 2.2.1, 3.1.8 (no (anchor) clause) and 3.2.2 (no (rsr-profile) clause). DEBT-REGISTER M-1 is marked superseded. src/Hypatia/ABI/RsrProfile.idr models only 3.2.2's profile resolution, which is unchanged, so it needs no mirror change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
…tia lacks Hypatia's structural_drift rule SD022 reads a backticked path as a path in this tree. src/assail/ is panic-attack's, so the ledger now names the module and the repository instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 36 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. |
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
5cba98e to
527a642
Compare
Signed-off-by: Jonathan D.A. Jewell <6759885+hyperpolymath@users.noreply.github.com>
|
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4d5538c8c6
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| defp descriptile_carrier do | ||
| dir = present_mr("descriptiles") | ||
|
|
||
| fn repo -> | ||
| if Enum.any?(@descriptile_clauses, &deed_clause?(repo, &1)), do: :pass, else: dir.(repo) |
There was a problem hiding this comment.
Wire 3.1.1 to the deed-aware detector
For every deed-era repository that has a top-level descriptile clause but no legacy descriptiles/ directory, this helper is never called because detectors/0 still maps 3.1.1 to present_mr("descriptiles"). Those repositories therefore continue to receive a false failure for 3.1.1; register descriptile_carrier() as the criterion's detector.
Useful? React with 👍 / 👎.
| record_files == [] -> | ||
| if deed == :ok, do: :pass, else: :fail |
There was a problem hiding this comment.
Check the deed when legacy record files are present
When at least one record-dialect file exists, both identical record_files == [] conditions are false, so execution only validates the legacy files and never considers the deed state. A repository with a malformed chora deed and a valid remaining .a2ml file consequently passes 3.2.1, contrary to the new every-carrier requirement and the added regression test.
Useful? React with 👍 / 👎.
| with the `+.a2ml+` as the fallback (A7 M5): 3.1.2-3.1.9 by clause, 3.1.1 by | ||
| the deed carrying any descriptile clause, and 3.2.1 by every carrier parsing, | ||
| the deed first. Against the bundled criteria hypatia now fails 1.2.4, 2.1.4, | ||
| 2.2.1, 3.1.8 (no `+(anchor)+` clause: the source anchor had no `+authority+`) |
There was a problem hiding this comment.
Remove 3.1.8 from the documented failure list
The repository's hypatia_chora.deed already contains a valid top-level (anchor) clause with an :authority, and anchor_present/0 recognizes that carrier, so the self-score passes 3.1.8. Listing it among the current failures makes this newly added debt status and its score summary inaccurate.
Useful? React with 👍 / 👎.
… (main does not compile) (#1128) ## What `main` has not compiled since #1117 (0cc4455). In `lib/rules/rsr_conformance.ex`: - `descriptiles_parse/1` calls `deed_state/1`, which nothing defines (`CompileError`). - The same `cond` has a duplicate `record_files == []` branch that can never run. - `descriptile_carrier/0` is defined but never used. That is an unused-function warning, and 3.1.1 still read only the directory. #1117's own head (4d5538c) has the same defect. It merged while its Test, Check and Compile jobs were still queued. This PR does what #1117's description says: - `deed_state/1`: the chora deed is `:absent`, `:ok` (it parses as a `repo-deed`) or `:broken`. `deed_parses?/1` is folded into it. - **3.1.1** uses `descriptile_carrier/0`: the deed first, then the `descriptiles/` directory. - **3.2.1**: a broken chora deed fails, even when the `.a2ml` files parse. ## Why Every open hypatia PR fails Compile and Test until `main` compiles. The owner asked for this fix. ## Evidence - #1117's own tests are on `main` but cannot compile there: "3.1.1: a chora deed with a top-level descriptile clause is the substrate", "3.1.1: a deed with none, or only nested, fails; the directory alone passes", "3.2.1 needs every carrier to parse, the deed first". - On `origin/main`: `(CompileError) … cannot compile module Hypatia.Rules.RsrConformance`. - On the code before #1117: `37 tests, 9 failures`, including those tests. - With this change: `37 tests, 0 failures`. - `mix test test/rules test/soundness_test.exs`: `974 tests, 0 failures`. - `mix format --check-formatted` passes on the changed file. `gen-rules-praxis.sh --check` and `check-readme-claims.sh` also pass. ## Not done - The full suite was not run. My local copy of the repo is partial, so CI runs the rest. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZkGCWGEyedTtdEHAPT)_ Co-authored-by: Claude <noreply@anthropic.com>
…ompiles again (#1125) ## Summary **hypatia main has not compiled since #1117 (`0cc44552`).** `lib/rules/rsr_conformance.ex` calls `deed = deed_state(repo)` at line 652 but never defines `deed_state/1`. The result is `undefined function deed_state/1` in `Hypatia.Rules.RsrConformance`. The cause is merge commit `4d5538c8` ("Merge branch 'main' into rsr/criteria-accept-chora-deed"), which #1117 was squashed from. It kept the new call but dropped four pieces: - the `deed_state/1` definition; - the 3.2.1 `deed == :broken -> :fail` branch; - 3.1.1's `descriptile_carrier/0`; - the deed-first anchor read for 3.1.8. #1117's own tests expect the dropped behaviour. ## Changes This touches `lib/rules/rsr_conformance.ex` only. It restores the file as #1117's last commit before that merge (`d0880bb5`) had it: - 3.1.1 reads a top-level descriptile clause of the chora deed; - 3.1.8 reads the `(anchor …)` clause first; - 3.2.1 fails a chora deed that does not parse, whatever `.a2ml` remains. ## Verification - `mix compile --warnings-as-errors`: clean. - `rsr_conformance` tests: 33 pass, 0 failures. Both were run with Elixir 1.19 / OTP 27. Split out from `rsr/scorecard-assessor` so that main is fixed first, without waiting on the scorecard change. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC)_ Signed-off-by: Jonathan D.A. Jewell <6759885+hyperpolymath@users.noreply.github.com> Co-authored-by: Claude <noreply@anthropic.com>
## Summary This is the writer half of the owner's 2026-11-14 ratification of the RSR scorecard deed. The grammar half is metadatastician/standards#1586 (`rsr-scorecard.abnf` v0.2.0). ## Changes - `RsrConformance.to_deed/2` writes: - `:assessor "hypatia rsr_score <version>"`, with the version from `mix.exs` at compile time; `assessor/0` returns the same string; - `:assessed-at`, the run time in UTC with whole seconds. - A new `:assessed_at` option lets tests inject a fixed `DateTime`. Offsets are converted to UTC. - Tests are extended. `docs/consumers/rsr-oracle.adoc` and the mix task docs are updated. **Includes #1125's fix (`89c0ab79`)**, which restores the part of #1117 lost in its merge, because main does not compile without it. Merge #1125 first; this branch then merges cleanly. ## Verification - `rsr_conformance`: 34 pass, 0 failures. - `mix format`: clean. `mix compile --warnings-as-errors`: clean. - `mix hypatia.rsr_score --write` on a copy of hypatia produces a deed that passes both `deed_lint --rsr-scorecard` and the plain linter, from the new standards branch. Merge after metadatastician/standards#1586. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC)_ --------- Signed-off-by: Jonathan D.A. Jewell <6759885+hyperpolymath@users.noreply.github.com> Co-authored-by: Claude <noreply@anthropic.com>
) ## Summary Main does not compile since #1117. `descriptiles_parse/1` calls `deed_state/1`, which nothing defines. A duplicate `record_files == []` branch could never run, and `descriptile_carrier/0` was never called. ## Changes - 3.1.1 now uses `descriptile_carrier/0` (deed first, then the directory), as #1117 describes. - `deed_state/1` replaces `deed_parses?/1` and returns `:ok`, `:none` or `:broken`. - 3.2.1: a broken chora deed fails even when the `.a2ml` files parse, as #1117 says. The dead duplicate branch is removed. ## Verification - `mix compile --warnings-as-errors`: clean (it failed before with the undefined function and the unused function). - `mix format --check-formatted` on the file: clean. - `mix test`: 3514 tests, 0 failures, 246 excluded (LANG=C.UTF-8). Not run: Rust, Zig and Idris2 checks. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZmUcth17dd8P9G2L1G)_ Signed-off-by: Jonathan D.A. Jewell <6759885+hyperpolymath@users.noreply.github.com> Co-authored-by: Claude <noreply@anthropic.com>
#1133 broke (#1136) ## Summary #1133 (Deed parser v2) merged while one Codex finding and four red checks were still open. Three of the four reds came from #1133 and are fixed here. The fourth was already red on main. ## Changes 1. **Codex, `lib/hypatia/deed.ex:104`.** `dispatch/1` accepted any `*_praxis.deed` without checking the stem. - In deed.abnf (146–148), `praxis-file` and `repo-file` share one rule: `stem = 1*( ALPHA / DIGIT / "-" / "." / "_" )`. One `stem/3` now enforces it for both suffixes. - Tests: `_praxis.deed`, `☃_praxis.deed`, `café_praxis.deed` and `a b_praxis.deed` are refused by both `dispatch/1` and `parse/2`. `my-proj.v2_x_praxis.deed` is accepted. The chora cases still refuse. 2. **Elixir Compile (dogfood toolchain).** `check-no-active-logtalk.sh` fails on the vendored fixture `test/fixtures/deed/standards/valid/rsr-template-repo_chora.deed:110`, which carries the upstream `:report-format "logtalk"`. - The fixture is added to `.logtalk-allowlist` with the same class ("vendored upstream fixture") and the same removal condition as its two identical siblings already listed. - The list's header asks for rule-level review of a new entry. That review belongs here, because editing the fixture would break its byte-faithfulness to standards. 3. **governance / UUID v7 and UUID v8 conformance.** `check-uuid-v8.sh --strict` (pinned at standards 7641c7f) flagged four ids in `test/deed_conformance_test.exs` that are deliberately not v8. - Those inputs are now built at run time from one v8 id by swapping its version and variant nibbles. The assertions are unchanged, and the v5 case now also checks the refusal reason. - The test file and three vendored invalid fixtures go into `.uuid-v8-exempt`, following the `test/deed_test.exs` precedent. The fixtures pass the pinned gate but fail the current standards script, so the exemption keeps them from breaking when the pin moves. - No real id or domain is touched. 4. **Escript packaging soundness: not this change.** It failed on `undefined function deed_state/1`, which came from #1117's merge. That red push run sat on 6033a78, before #1125/#1128/#1132. The same error reproduces on 6033a78, and an unrelated PR failed the same way before #1133 merged. Current main passes (43638218845). ## Verification Run with Elixir 1.19.4 / OTP 27.3.4.19, the CI versions: - `mix compile --warnings-as-errors`: exit 0. - `mix test`: 3338 tests, 0 failures, 246 excluded (seed 0; `verisim_data` and the other excluded tags as configured). - `test/soundness/run-escript-soundness.sh`: 14/14. - The consumer-shaped scan output passes `check-findings-schema.sh`. - `mix format --check-formatted` passes on the changed files. - `check-uuid-v8.sh --strict` passes at the pinned and the current standards versions. The logtalk guard reports 15 allowlisted, 0 violations. ## For the owner - **Gate looser than the grammar:** the standards gate `deed_lint.js` (702–703) doesn't check the praxis stem, so it accepts names the grammar refuses. The reader now follows the grammar. - **Formatting on main:** 27 files on main fail `mix format --check-formatted` under 1.19.4, and no workflow runs that check. Not touched here. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC)_ --------- Co-authored-by: Claude <noreply@anthropic.com>



Summary
This PR is stacked on #1113. On 2026-11-14 the owner ruled that RSR criteria 3.1.x accept the chora deed, with
.a2mlonly as the A7 fallback.Changes
descriptiles/directory still counts..a2mlfiles parse..a2ml. Their verdicts don't change.RsrProfile.idrmodels only 3.2.2, so it needs no change.rsr_conformance_test.exs: the#u5test ids become#u8/#uuid(owner ruling Q2). DEBT-REGISTER M-1 is marked superseded.Verification
rsr_conformancetests: 30/30.mix test: 3225 tests, 10 failures, all of them theHttpcRedirectCredentialsTestfailures that main also has.Not done here (owner decision)
The canon
0-canon/rsr/rsr-criteria-v2.a2mlstill names only.a2mlcarriers in 3.1.1–3.2.2.Changing that text would add a fact to a retired A2ML file, and
canon.lockpins the file by hash. A7 §10 instead routes the canon to a criteria praxis deed generated by kyX. This PR therefore changes the oracle and leaves the canon alone.Generated by Claude Code