agentic-praxis-grimoire

ADR 0010: Six-Skill Post-Superpowers Stability Dispositions

Status

Accepted

Date

2026-07-20

Acceptance authority

The human maintainer’s APG13 assignment accepts this decision. It authorizes individual maturity disposition, at most one demonstrated material procedure correction per skill, current-document reconciliation, one private development commit and push, and required managed reports. It does not authorize a public candidate, publication, active integration change, external-repository mutation, Superpowers restoration, or APG14 work.

Context

APG12A is accepted as complete: it corrects public lineage and read-only validation while preserving ADR 0009, version-1 schemas, command surfaces, skills, projections, maturity, public and reference repositories, RepoMap, and active integration. Its corrected terminal suites contain 15 checker unit and 58 checker integration tests, 8 release unit and 32 release integration tests, 8 user unit and 39 user integration tests, 22 report-tool Bats families, and 28 project-skill integration families.

ADR 0006 targets all six current skills for individual stable disposition. It makes repeated real use, representative non-triggers, absence of unresolved material authority/privacy/safety/procedure defects, post-Superpowers review, and supported rollback mandatory. Clean A/B superiority, automatic invocation, publication, global installation, and positive use in every second repository are not independent blockers.

Decision

Stable criteria

Each skill is reviewed for:

  1. one coherent owner with precise positive and non-trigger boundaries;
  2. preservation of human, task, and repository authority;
  3. a coherent, proportional, independently removable procedure;
  4. repeated real use, non-trigger and edge/stop evidence, and fresh completion evidence;
  5. safety, privacy, destructive-action, source-trust, and publication boundaries;
  6. post-Superpowers positive or direct-review evidence without retired-tool dependency;
  7. correction history, regression status, maintenance, and rollback; and
  8. current mechanical regression plus fresh non-author maturity review.

stable means suitable for routine bounded use within the recorded trigger and project boundaries. It does not mean universal applicability, automatic invocation, comparative superiority, production warranty, or independent action authority.

Evidence classes

Positive real use, textual scenario application, non-trigger evidence, integration evidence, mechanical evidence, distribution evidence, and historical correction are recorded separately. Projection, installation, release, discovery, and catalog success do not become positive semantic use. Non-triggers do not count as positive use.

Frozen current applications

Before any maturity change, APG13 froze one positive, non-trigger, and edge or stop family per skill. Fresh read-only workers applied the unchanged leaves:

All six applications recommended stable-candidate. None demonstrated a material defect.

Corrections and rollback

No procedure correction was made. Every SKILL.md remains byte-identical to the APG12A baseline, so each APG13 correction count is zero. If a maturity disposition becomes unsupported, restore that catalog row to provisional, supersede the disposition, reconcile current owners, and rerun affected gates without erasing historical evidence. Behavior-bearing rollback remains owned by the skill lifecycle and each skill’s recorded correction history.

Individual dispositions

Skill Evidence summary Final non-author review Maturity
composing-bounded-worker-assignments APG4 scenarios; repeated APG6-APG12A bounded assignments; direct-work non-triggers; overlap/authority stops; post-retirement use accept-stable stable
designing-significant-changes APG4 scenarios; APG6 RepoMap and APG7/APG11/APG12 design; multiple accepted-design non-triggers; reversible/reserved-choice edge accept-stable stable
planning-repository-work APG4 scenarios; APG6-APG12A dependent plans; one-action non-triggers; invalidated-plan stops accept-stable stable
implementing-with-test-discipline APG4 scenarios; APG7/APG7A/APG10-APG12A behavioral work; APG8/APG9 and documentation non-triggers; APG10 cleanup correction and ownership stop accept-stable stable
debugging-systematically APG4 scenarios; APG7A/APG11-APG12A diagnosis; APG9 and expected-red non-triggers; APG11 no-production-fix and stale/authority stops accept-stable stable
reviewing-and-verifying-repository-work APG4 scenarios; APG5-APG12A review; casual-feedback non-trigger; APG6 frontmatter correction; stale/missing-authority stops accept-stable stable

All six catalog rows become stable. LT21 and LT24 become implemented-by-APG13.

Release boundary

APG13 creates no real v0.2.0 candidate and performs no publication or active integration update. Public v0.1.0 retains its original commit, tree, lightweight tag identity, historical provisional catalog, and role as the active public-backed source. APG14 is ready under the v0.2 objective but remains separately authorized and unstarted.

Alternatives considered

Promote all six from accumulated evidence without fresh review

Rejected. Historical strength does not replace frozen current applications, fresh resulting-state evidence, or non-author disposition.

Require clean A/B superiority

Rejected as a mandatory gate. It may be useful future evidence but is not a material defect under ADR 0006.

Require second-repository positive use for every skill

Rejected as a universal gate. Cross-repository breadth is recorded where present; repeated positive APG use may satisfy maturity.

Retain every skill provisional indefinitely

Rejected. It would disregard the accepted criteria and accumulated evidence without naming a concrete defect.

Review and disposition each skill separately

Accepted. Evidence distribution, correction history, edge behavior, and rollback differ by owner.

Treat distribution success as maturity

Rejected. Distribution proves availability and integrity, not semantic procedure quality.

Treat only a concrete material defect as a blocker

Accepted. Stylistic preferences, hypothetical failures, and absent non-required claims do not consume correction budgets or block promotion.

Consequences

Deferred decisions