EPIC: v0.2 maintenance roadmap (corrective · adaptive · perfective)

[hyperpolymath]

Durable backlog for arghda-core v0.2, organized by ISO/IEC-14764 maintenance type × must/intend/wish. Every item ladders up to the project's two north stars: trust (a proven/ file is actually still proven — kill the silent-failure class) and reusability across provers (the language-agnostic engine consumed by arghda-panll / arghda-studio).

This issue stands in for a GitHub Milestone (none could be created via the available tooling).

Corrective — faults relative to the goal

• corrective/must: transitive-import closure hashing for proven staleness #12 · must · transitive-import closure hashing for proven staleness (in feat(proven): transitive-import closure hashing + cycle-safety regression #26)
• corrective/must: lock in cycle-safety of the import walk with a regression test #13 · must · lock in cycle-safety of the import walk with a regression test (in feat(proven): transitive-import closure hashing + cycle-safety regression #26)
• corrective/intend: parser resilience + negative/property tests #14 · intend · parser resilience + negative/property tests
• corrective/intend: known-answer tests for the custom timestamp.rs #15 · intend · known-answer tests for the custom timestamp.rs

Adaptive — adapt to provers / toolchain / estate / consumers

• adaptive/must: make the "language-agnostic" claim honest (Agda-first today) #16 · must · make the "language-agnostic" claim honest (Agda-first today)
• adaptive/must: wire the standards rsr-profile capability check into CI #17 · must · wire the standards rsr-profile capability check into CI
• adaptive/intend: introduce the Language abstraction (multi-prover) + a second backend #18 · intend · introduce the Language abstraction (multi-prover) + a second backend (headline epic; 0.2 breaking)
• adaptive/intend: declare MSRV + rust-toolchain.toml + CI matrix #19 · intend · declare MSRV + rust-toolchain.toml + CI matrix
• adaptive/intend: Groove protocol endpoint for PanLL discovery #20 · intend · Groove protocol endpoint for PanLL discovery

Perfective — quality / performance / evolvability

• perfective/must: version the JSON contracts + adopt SemVer #21 · must · version the JSON contracts + adopt SemVer (do before panll/studio depend on them)
• perfective/intend: incremental scan (cache by content-hash) #22 · intend · incremental scan (cache by content-hash)
• perfective/intend: widen the config surface #23 · intend · widen the config surface
• perfective/intend: reporting modes (--format pretty|json|sarif) #24 · intend · reporting modes (--format pretty|json|sarif)
• perfective/wish: parallel linting, metrics, arghda-studio #25 · wish · parallel linting, metrics, arghda-studio

Recommended sequence

1. corrective/must: transitive-import closure hashing for proven staleness #12 → 2. corrective/must: lock in cycle-safety of the import walk with a regression test #13 (both in feat(proven): transitive-import closure hashing + cycle-safety regression #26) → 3. adaptive/must: make the "language-agnostic" claim honest (Agda-first today) #16 + adaptive/must: wire the standards rsr-profile capability check into CI #17 → 4. perfective/must: version the JSON contracts + adopt SemVer #21 → then 5. the adaptive/intend: introduce the Language abstraction (multi-prover) + a second backend #18 Language epic for 0.2.

(Preventive-flavored items — cycle guard #13, timestamp KATs #15, contract versioning #21 — are slotted by dominant character; say the word for a separate preventive lane.)

[hyperpolymath]

Parked governance changes (maintainer-decided) — record only, NOT yet executed

Two RSR governance changes are decided but PARKED pending a canonical-governance-store rework. Recorded here as planned-but-parked; the estate-wide edit is not to be executed until the rework lands. Revisit A + B together post-rework.

Item A — add preventative to Axis 2 (RSR maintenance-type axis)

• New order: corrective > preventative > adaptive > perfective (preventative slots in at feat: finish the v0.1 engine — dag, check, state machine, event stream #2).
• Rationale (ISO/IEC 14764): corrective + preventative are the fault-oriented pair (realized vs latent faults); adaptive + perfective are improvement-oriented. Fault work outranks improvement, so preventative goes at feat: finish the v0.1 engine — dag, check, state machine, event stream #2.
• Definition to add: "Preventative — proactively detect and remove latent faults, fragility, and tech-debt (hardening, defensive refactors, STUB/PARTIAL shoring-up, dependency/security pre-emption) before they escalate into corrective work; sourced from the honest Axis-1 state and Axis-3 audit findings."
• Blast radius — must move in LOCKSTEP: the literal string corrective > adaptive > perfective appears in ~30 files across 7 repos (krl, nextgen-typing, tangle, quandledb, nextgen-databases + sub-projects): in .machine_readable/META.a2ml, .machine_readable/policies/MAINTENANCE-AXES.a2ml, the governance .adoc/.a2ml docs, and the validate-rsr gate (Justfile) which greps the literal string. The META value and the gate's expected string must change together or validate-rsr breaks estate-wide.
• This EPIC's own title triad (corrective · adaptive · perfective) is part of that lockstep and is retitled to include preventative at the same time — not changed now (parked).

Item B — Axis-1 vocabulary normalisation

• Canonical Axis-1 lowest bucket is like (must > intend > like).
• The 2026-06-19 per-repo scoped tables (docs/identity-fabric/musts-intends-wishes.adoc) currently say "Wish" — normalise Wish → like when the rework lands.

Status

PARKED by the maintainer — canonical governance store under rework. Do not execute the estate-wide lockstep edit yet; revisit A + B together post-rework.

Cross-ref: hyperpolymath/nextgen-typing#69.

---

Generated by Claude Code