Restore RFC 008 manifest contracts - #1110
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
There was a problem hiding this comment.
Alignment Review Report
Two-tier review of the RFC 008 manifest-contract restoration. Net finding: this is a byte-for-byte restore of the already-reviewed-and-merged slice 0a (#1044) contract files, plus a docs line. All PR-owned files pass every automated check; no mechanical issues.
Automated Checks
- Lint: PASS (this PR's files).
ruff format --check,usort check, andruff check src/ tests/all pass on the changed paths (8 files already formatted).bash .claude/hooks/lint.shdoes exit non-zero, but every file it flags is a pre-existingenvs/*formatting issue that this PR does not touch. - Debug code: CLEAN (this PR's files).
bash .claude/hooks/check-debug.shonly reports pre-existingprint/console.print/TODOinsrc/openenv/cli/...; nothing insrc/openenv/validation/. - Tests: 23/23 pass in
tests/test_validation/;sync_validation_schemas.py --checkreports in sync;import openenv.validationresolves (17 exports); no dangling imports to the non-reapplied submodules;tests/test_validation+tests/test_clicollect cleanly (154 tests).
Open RFCs Context
- RFC 008 — Environment Auto-Validation (Status: In Review, author @zkwentz). This PR implements the RFC's "PR2 — contracts" slice (normalized manifest schema, core enums, committed JSON Schema + CI sync). The RFC's 44-test mapping is annotated "Approved 2026-08-04".
- RFC 004 — Rubrics (referenced, not modified) — the
llm_judged/rubric_treecapability fields leverage RFC 004 as the served-format judged path.
Tier 1: Fixes Required
None. The lint/debug hook findings are all pre-existing and in files unrelated to this PR.
Tier 2: Alignment Discussion
Principle Conflicts
None identified. The manifest only declares the reward range/tolerances for validation — it does not compute or externally augment rewards — so the "Rewards in environment" invariant (INVARIANTS.md, authored by @Darktex) is respected. No client↔server imports, no Gymnasium API surface changes, no MCP/reset exposure.
RFC Conflicts
One informational flag (low-risk, not blocking):
ALIGNMENT FLAG: Implements contracts from an In-Review RFC
- Principle/RFC at stake: RFC 008 (Status: In Review, not yet Accepted)
- The concern: The two-tier model flags changes that implement an unfinalized RFC so the design discussion stays focused. Mitigating context that lowers the risk here: (1) the reapplied files (
types.py,manifest.py,__init__.py,manifest.schema.json, sync script, fixtures/tests) are verified byte-identical to the previously-merged #1044; (2) this restores only the pure Pydantic contracts — it deliberately does not re-add the CLI wiring / runner / report / graders from later slices, so it avoids thefastmcpcompatibility concern that motivated the revert in #1108. Worth a maintainer confirming the reapplication scope matches intent. - Suggested reviewer: @zkwentz (RFC 008 + #1044 author)
One optional inline nit on variance_tolerance (see inline comment).
Summary
- 0 mechanical issues to fix
- 1 alignment point for human review (informational: implements an In-Review RFC)
- 0 blocking RFC conflicts; 0 principle conflicts
Sent by Cursor Automation: Pre-review


This PR reapplies #1044 after #1108. Stack: #1110 → #1111 → #1112.