Merge pull request #82 from egohygiene/codex/continuity-skill-integration

feat: integrate repository continuity across agent workflows

Author: szmyty

CHANGELOG.md

@@ -10,6 +10,10 @@ This repository follows release-tag based versioning for install pinning and rel
 
 ## [Unreleased]
 
+- Integrated repository continuity across every canonical first-party skill
+  and agent through a reviewed disposition inventory, compact composition
+  contracts, managed Codex/Copilot/Claude instruction projections, and
+  deterministic resume/completion lifecycle evaluations (#80).
 - Added the provider-neutral repository continuity specification, schema,
   bounded root Markdown template, maintenance skill, prototype migration
   report, deterministic fixtures, and downstream integration map (#79).

README.md

@@ -1,6 +1,8 @@
 # Aether
 
-Aether is Ego Hygiene’s canonical first-party library of reusable AI specifications, skills, and agent source with deterministic validation and build tooling.
+Aether is Ego Hygiene’s canonical first-party library of reusable AI
+specifications, skills, agent source, and instruction modules with
+deterministic validation and build tooling.
 
 Architecture navigation: [META.md](META.md) inventories Aether's complete
 18-document architecture graph. The reusable materialization contract, schema,
@@ -18,7 +20,8 @@ and consumer validator live under
 ## 2) What Aether owns (and does not own)
 
 Aether owns:
-- First-party canonical specs, skills, and agents.
+- First-party canonical specs, skills, agents, and reusable instruction
+  modules.
 - Catalog/provenance schemas and deterministic validators.
 - Distribution/projection build scripts.
 
@@ -33,11 +36,14 @@ Canonical source:
 - `library/organization/specs/`
 - `library/organization/skills/`
 - `library/organization/agents/`
+- `library/organization/instructions/`
 
 Generated artifacts:
 - `catalog/first-party/catalog.v1.json`
 - `dist/skills/`
 - `dist/github/`
+- `dist/codex/`
+- `dist/claude/`
 
 `.staging/` is non-canonical migration/provenance holding space and cannot be emptied without ADR-005 deletion-gate requirements (`DECISIONS.md#adr-005`).
 
@@ -47,6 +53,7 @@ Current canonical inventory snapshot (recompute with the commands shown):
 - 29 specifications (`find library/organization/specs -name "*.spec.md"`)
 - 34 skills (`find library/organization/skills -name "SKILL.md"`)
 - 9 canonical agent profiles (`find library/organization/agents -name "AGENT.md"`)
+- 1 canonical instruction module (`find library/organization/instructions -name "INSTRUCTION.md"`)
 
 Machine-readable catalog and provenance:
 - `catalog/first-party/catalog.v1.json`
@@ -151,6 +158,7 @@ Inspect generated output:
 ```sh
 find dist/skills -maxdepth 2 -type f | sort
 find dist/github -maxdepth 4 -type f | sort
+find dist/codex dist/claude -maxdepth 4 -type f | sort
 ```
 
 Validate publishability (no release write):

catalog/README.md

@@ -22,6 +22,9 @@ provenance, lifecycle, and distribution.
 - `schemas/aether.repository-continuity.v1.schema.json` — root operational
   handoff metadata, base/candidate/live state, review evidence, size, privacy,
   and stale/superseded lifecycle contract
+- `schemas/aether.continuity-dispositions.v1.schema.json` — complete canonical
+  skill/agent reader-writer, reader, read-only, and not-applicable composition
+  inventory
 - `first-party/catalog.v1.json` — canonical compatibility catalog for the current specification and skill corpus
 - `external/approved-skills.v1.json` — governed external skill catalog entries reconstructed from staged provenance
 - `external/source-candidates.v1.json` — non-publishable external source candidates, including their rights-review state
@@ -40,6 +43,8 @@ Current source-specific catalogs have different responsibilities:
 
 - specification and skill frontmatter is authoritative for `first-party/catalog.v1.json`;
 - canonical agent source and `library/organization/agents/catalog.json` own agent-specific capability metadata;
+- `library/organization/instructions/repository-continuity/` owns the reusable
+  continuity pointer and reviewed per-skill/per-agent dispositions;
 - reviewed external source records own their upstream provenance and redistribution evidence.
 
 The normalized provenance model is a **derived policy view**, not a second hand-edited source of truth. It gives these artifact kinds one common vocabulary for source, revision, digest, license, trust, lifecycle, compatibility, maintainer ownership, and publication readiness.
@@ -71,7 +76,10 @@ The normalized provenance contract supports:
 - prompts;
 - instructions.
 
-The current repository has canonical specs, skills, and agents. Prompt/instruction source can be added later without inventing a separate provenance lifecycle. Rich agent catalog/cost-control work remains separate from the generic provenance contract.
+The current repository has canonical specs, skills, agents, and a repository
+continuity instruction module. Future prompt/instruction sources can use the
+same provenance lifecycle. Rich agent catalog/cost-control work remains
+separate from the generic provenance contract.
 
 ## First-party and external separation