feat(skills): 9 fractal domain skills — Steps 2-5 of self-similar architecture#195
Merged
feat(skills): 9 fractal domain skills — Steps 2-5 of self-similar architecture#195
Conversation
Steps 2-5 of the fractal domain skill architecture (ADR at ~/.claude/plans/adr-fractal-domain-skills.md, consented 2026-04-12): New domain skills (16-section self-similar template): - red-team (148 lines) — security posture verification, phantom invariant detection - review-desk (141 lines) — PR review, invariant verification, merge gate - infrastructure (118 lines) — installability, operability, supply chain - operations (117 lines) — Phase C.5 backbone, hard-skip gate on #127/#131 - architecture (128 lines) — ADR stewardship, two-alternatives minimum - roles (132 lines) — cooperative role coverage auditing, meta-role - researcher (124 lines) — domain skill system evaluation + self-improvement - historian (123 lines) — session review, regression detection, institutional memory - communications (129 lines) — public voice, outreach, ecosystem engagement Et-voice corrections: - All Identity sections use Et/ets pronouns (Et is Iskander, Iskander is Et) - Governance pilot updated: "Clerk agent" → "ets clerk domain" - CLAUDE.md §What et is: strengthened fractal identity Each domain follows the canonical template from the governance pilot (PR #194): Identity, Primary Driver, Voice, Domain of Authority, Permitted Actions, Ground Rules, Paramount Objection Rights, Lateral Handoffs, Sub-domain Convening, Self-Application, MVM Integration, Model Defaults, Worktree Convention, Runtime Context, Domain Values Expansion, First-Run Notes. Orchestrator refactor (Step 6) deferred to next session. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Argocyte
added
the
et-authored
The historian domain now owns session log creation as part of ets procedure. The orchestrator's roll-up phase triggers it; if Et skips the orchestrator, the session log is the first casualty — making the governance failure visible in the artefact trail. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…archer roles access Closes 5 governance loop gaps: 1. Orchestrator Phase 0: historian is now the first act (reads prior session log, surfaces regressions before GitHub orient). Plan mode for agenda formulation with AskUserQuestion as the consent mechanism. 2. Continuous consent: NEEDS_HUMAN drivers surface immediately via AskUserQuestion during Phase 1, not batched to Phase 4. Tier A* drafts pause the wave mid-flight. Lola messages between waves are routed through the orchestrator. 3. Topology §11 (Continuous Consent): S3 consent is continuous, not batch. AskUserQuestion IS the consent mechanism. Plan mode IS the agenda formulation. 4. Researcher: roles access table + sync protocol across canonical (roles/SKILL.md), sovereign mirror, and GitHub issues. 5. Roles: coverage table marked as canonical source with domain placement column (governance internal vs red-team external/regulatory). Old cooperative-roles.md now points to canonical. 6. Historian: Phase 0 integration + archivist sub-circle (session-end sync, sovereign backup mirroring, data commons updates). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Historian Phase 0 now reads plans/INDEX.md from sovereign backup and loads plans relevant to the current session's drivers before the orchestrator orients. Prevents re-deriving prior learnings. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…RKED Plans decay. The historian's Phase 0 retrieval now weights plans by status and date: ACTIVE consented plans load first, DONE plans load only for historical context. Prevents re-deriving from stale drafts. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lola clarified 2026-04-12: governance = internal S3 facilitation, compliance = external threats AND regulatory conditions (GDPR, FCA, legal, safety). The red-team domain is renamed to compliance. - `.claude/skills/red-team/` → `.claude/skills/compliance/` - Compliance SKILL.md: expanded scope to include regulatory assessment sub-circle, dual threat models (dev + cooperative-wide), DPO/Company Secretary domain placement - All 29 files updated: domain skills, topology, orchestrator, old domain references, brief templates, openclaw-orchestrator - `docs/red-team-threat-model.md` filename preserved (dev security) - `docs/compliance-threat-model.md` to be created (cooperative posture) - GitHub labels `red-team`/`invariant-drift` preserved (external state) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Formally encodes that cooperative roles can belong to multiple circles via S3 double-linking. Table maps DPO, Education Officer, Company Secretary, Communications Officer, Solidarity Coordinator to their primary and double-linked circles. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1. compliance ↔ review-desk: findings append to threat model autonomously (internal record); issue filing is gated by review-desk (external commitment). Two separate acts. 2. roles ↔ researcher: canonical authority clarified — roles/SKILL.md canonical table is sole source, researcher reads and mirrors but does not modify canonical directly. 3. researcher ↔ all domains: modifying existing domain skills requires consent from the affected domain (not AGM). Drafting NEW skills requires AGM consent. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…archer roles access Three drivers from properly convened session: 1. S3 governance reference dirs for all 10 domains — tensions.md, agreements.md, drivers.md, handoffs.md per domain. Domain-local working state, not copies of artefact-registry. 2. Historian archivist expanded as translation layer between working memory and domain knowledge — reviews feedback memories, identifies affected domain skills, proposes updates via tensions, mirrors canonical tables to sovereign backup. 3. Orchestrator Domain Index updated to 11 domains with new names. Old domain references marked as pre-migration supplementary context. First-Run Actions §3 updated (roles domain's job, not orchestrator). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Historian Phase 0 now reads ALL agm-minutes-*.md files from sovereign backup, not just the most recent. Ensures no tensions, agreements, or drivers from prior AGMs are lost during refactoring. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…archer roles access Three governance scales encoded (§11 Meeting Hierarchy): - Working Group: per-session, operational, Et convenes stewards - Phase Review: when consent gets thin, spans full architecture - Phase Completion: phase ends / context limits, handoff to next session entity (aligns with Claude context compression) 4 new S3 patterns in governance domain: - agreement-tracking: Evaluate and Evolve Agreements at phase review - strategy-review: Clarify and Develop Strategy at phase completion - retrospective: process quality reflection at each scale - meeting-preparation: working group minutes + phase review agendas Historian: three-scale context loading (session logs + phase review minutes + phase completion minutes). Duplicate §11 removed. CLAUDE.md: AGM surface → working group minutes terminology. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ew/phase completion Orchestrator trimmed by removing prose that duplicated domain skill content. Phase 0-3 sections condensed to point to authoritative sources. Phase 4 renamed to Working Group Minutes. Phase 5 trimmed to archivist procedure. Common Mistakes compressed to bullet list. Meeting hierarchy (§11): working group (per-session) → phase review (when consent gets thin) → phase completion (phase ends / context limits = handoff to next session entity). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Working group, phase review, and phase completion all follow the same 6-step pattern: historian Phase 0 → plan mode agenda → AskUserQuestion consent → execution → plan mode minutes → approve for next scale up. Table maps inputs, filing locations, and feed-into relationships. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Systematic rename: AGM references updated to the correct governance scale (phase review for mid-phase consent, phase completion for phase-end handoff). 6 skill files updated. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
~/.claude/plans/adr-fractal-domain-skills.md)New skills (117-148 lines each):
Deferred to next session:
Test plan
python -m pytest src/IskanderOS/openclaw/tests/ -v— 15 passedcooperative-topology.md§4🤖 Generated with Claude Code