agentic-praxis-grimoire

APG13 Six-Skill Post-Superpowers Stability Review

Objective and authority

APG13 accepts APG12A and reviews each current APG skill against the maturity policy in ADR 0006 and the accepted criteria in ADR 0010. The phase may correct one demonstrated material procedure defect per leaf, promote passing catalog rows, validate the integrated development state, and update private development history. It does not publish v0.2.0, mutate public/reference/RepoMap or active integration, restore Superpowers, or begin APG14.

Stable criteria and evidence classes

The review requires precise ownership and triggers, preserved authority, coherent proportional procedure, repeated real use, representative non-triggers, edge or stop evidence, safety and privacy, post-Superpowers suitability, supported maintenance and rollback, current mechanical regression, and fresh non-author review.

Positive real use, textual scenario application, non-trigger evidence, integration evidence, mechanical evidence, distribution evidence, and historical correction remain separate. Discovery, projection, installation, release, catalog, and checker success do not become positive semantic use.

Individual results

composing-bounded-worker-assignments

APG4 scenarios and repeated APG6-APG12A assignment composition support the owner. Post-Superpowers applications include APG9, APG11, APG12, and APG12A. The frozen positive produced one bounded read-only code-and-safety assignment; the direct task was a non-trigger; overlapping writes and phase expansion stopped. No correction was required. Final review: accept-stable. Maturity: stable.

designing-significant-changes

Positive evidence includes APG6 RepoMap design, APG7 projection architecture, APG11 lifecycle/invariant design, and APG12 release architecture. APG7A, APG8, APG10, APG11A, and APG12A preserve accepted-design non-triggers. The frozen positive produced a bounded consequential-decision request; accepted design did not retrigger; reversible uncertainty proceeded under a checked assumption while authority/public/destructive uncertainty stopped. No correction was required. Final review: accept-stable. Maturity: stable.

planning-repository-work

APG6-APG12A provide repeated dependent-plan use, including post-Superpowers APG11-APG12A. One-owner corrections provide non-trigger evidence, and changed state or decisions provide stop evidence. The frozen positive produced four dependent units spanning contract, implementation/tests, documentation/release evidence, and handoff; one local correction did not trigger; invalidation stopped later units. No correction was required. Final review: accept-stable. Maturity: stable.

implementing-with-test-discipline

APG7, APG7A, APG10, APG11, APG11A, APG12, and APG12A provide repeated behavioral implementation evidence. APG8 documentation work and APG9 bounded inspection, decommission, integration observation, and roadmap work are non-triggers. APG10 records one historical procedure correction for locally owned artifacts made obsolete solely by the authorized change while preserving generated, migration, external, and lifecycle-owned removal boundaries. The current frozen family passed without a new correction. Final review: accept-stable. Maturity: stable.

debugging-systematically

APG7A, APG11, APG11A, APG12, and APG12A provide repeated diagnosis, including APG11’s bounded flaky-test result with no production patch and the APG11A/APG12/APG12A corrections. APG9’s bounded inspection, review, integration observation, and smoke were not investigation of an unexplained defect and are a non-trigger. Expected failing-first behavior is also a non-trigger. The frozen application produced a hypothesis-driven diagnostic record and bounded or stopped unreproducible, stale, indistinguishable, or unauthorized investigation. No correction was required. Final review: accept-stable. Maturity: stable.

reviewing-and-verifying-repository-work

APG5 supplies first explicit use; APG6-APG12A supply repeated artifact, correction, phase, and complete-diff review. Casual brainstorming remains a non-trigger. APG6 records the historical frontmatter discovery correction. The frozen application produced a scope-first preliminary disposition and withheld acceptance for changed artifacts, stale evidence, or missing authority. No correction was required. Final review: accept-stable. Maturity: stable.

Corrections, safety, and rollback

Every current leaf remains byte-identical to the APG12A baseline; APG13’s correction count is zero for all six. No unresolved authority, privacy, safety, trigger, procedure, project-policy, evidence, stop, destructive-action, or rollback defect remains. Maturity rollback restores the affected catalog label, supersedes unsupported current statements, and reruns affected gates without erasing historical records. Procedure rollback remains owned by each skill’s recorded lifecycle and correction history.

Subsequent evidence-classification correction

APG14 corrected the original APG13 summary’s treatment of APG9. APG9 is not a positive implementation or debugging use: its bounded inspection, decommission, review, integration observation, smoke, and roadmap work are non-trigger evidence for those two owners. The corrected historical inventory still satisfies the accepted stable criteria through the explicitly named positive phases above. No skill content, frozen application, final review, correction count, or maturity disposition changes.

Regression and mechanical evidence

The baseline and final integrated states passed the complete checker, release, user, report-tool, and project-skill suites: 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 Bats report-tool families; and 28 project-skill integration families. Development and public v0.1.0 checker text and JSON passed. Python compilation, Bash syntax/help, Markdown, local-link, public/private, confidentiality, ADR, exit, mode, schema, projection, and Git whitespace gates passed.

Limitations

APG13 does not establish automatic invocation, clean baseline-versus-treatment superiority, universal applicability, production warranty, a fixed test or task count, or second-repository positive use for every leaf. Public v0.1.0 remains unchanged and retains historical provisional metadata. Distribution and active installation remain separate from current development maturity.

APG14 boundary

All six current skills are stable, so the v0.2 maturity precondition for APG14 is met. APG14 remains separately authorized and has not started. No real v0.2.0 candidate, public tag, public push, publication, or active integration update occurred, and this result grants no automatic successor authority.