Governed composition: Héphaïstos forges, capability registry, two gates#53
Draft
ifanjuang wants to merge 7 commits into
Draft
Governed composition: Héphaïstos forges, capability registry, two gates#53ifanjuang wants to merge 7 commits into
ifanjuang wants to merge 7 commits into
Conversation
Define the blueprint-in-Pantheon and adapter-outside model: where tool-specific templates and configurations live (OpenWebUI, Hermes, Langfuse and others) and the four disciplines that keep them adapted to Pantheon without coupling Pantheon to any tool. https://claude.ai/code/session_01UDCR8Dmp16xpFnwqJwsrPH
Resolve Issue #25 on the open #26 branch (consolidated to keep a single editor of the governance index files): - abstract product names in MODULAR_DOMAIN_REORIENTATION body, tables and diagram; confine names to the bindings registry with an explicit bindings/adapters exception; - reconcile the module manifest status with MODULE_ACTIVATION (status, activation, task_authorization as three axes); - add the reconciles-not-replaces hierarchy note; - frame a domain pack as a governed methodology configuration, not a runtime module; - index MODULAR_DOMAIN_REORIENTATION, ADAPTERS_AND_BINDINGS, CAPABILITY_PLACEMENT and the orphan docs (architecture pack, knowledge ingestion, workflow lifecycle) in STATUS, README, MODULES and CORE_CONCEPTS_MAP; - flag DATA_PLATFORM_* as to-verify pending a CLAUDE.md boundary review. AgentOS distillation (#27) left out of scope. https://claude.ai/code/session_01UDCR8Dmp16xpFnwqJwsrPH
Doc-only boundary review per #30, folded into the open reconciliation branch (same index files, refines #26's own edits): - DATA_PLATFORM_ARCHITECTURE: Directus no longer 'controls'; deployment profiles reframed as external candidates with no-authorization disclaimer; table families marked conceptual, not approved schema; adapter/binding status note added; - realign ARCHITECTURE_AGENCY/KNOWLEDGE_INGESTION/WORKFLOW_LIFECYCLE to candidate/to-verify in STATUS, README, MODULES to match their headers; - DATA_PLATFORM_* stay to-verify; nothing promoted. https://claude.ai/code/session_01UDCR8Dmp16xpFnwqJwsrPH
Add a governance architecture for forging workflows on the fly: - CAPABILITY_REGISTRY.md: capabilities declared as a dependency graph, metadata only, the index Héphaïstos forges from. - WORKFLOW_SCHEMA.md: governed-composition section (forge, retrieve/ reuse/revise/retain loop, two gates, per-step signatures; forged != authorized). - reference_reviews for Voyager/DSPy (forge runtimes, rejected as architecture) and EviBound/SkillsVote/GovernSpec/MedSkillAudit (convergent gate vocabulary). - index updates, CHANGELOG, ai_log. Distilled from external work; imports no engine. Execution stays external. https://claude.ai/code/session_01UDCR8Dmp16xpFnwqJwsrPH
…nto current The remote branch was last updated 2026-05-31; all of its content has since been incorporated into main via later merges. Keep current (newer) tree.
AGENTS.md and GOVERNANCE_COLLEGE.md now state explicitly that HÉPHAÏSTOS forges Workflow Manifest candidates (composing declared capabilities into a recipe for the cap), backing the governed-composition section of WORKFLOW_SCHEMA.md. A forged recipe stays a candidate; forging does not authorize execution. https://claude.ai/code/session_01UDCR8Dmp16xpFnwqJwsrPH
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
2 participants
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.
Intent
A governance architecture for forging workflows on the fly without duplicating the execution runtime. Distilled from external work, importing none of it.
What this adds
CAPABILITY_REGISTRY.md(new) — capabilities declared by governance metadata only, organized as a dependency graph (the executable lives in the runtime). It is the index HÉPHAÏSTOS forges from. A declaration is a candidate until reviewed; enrichment is governed; superseded declarations are archived (CHARON), not deleted.WORKFLOW_SCHEMA.md— new Governed composition section: HÉPHAÏSTOS forges a Workflow Manifest candidate for a specific cap; a retrieve / reuse / revise / retain loop mapped onto existing governance; two gates (pre-execution eligibility arbitrated by ZEUS, post-execution evidence verification); per-step signatures as governance contracts.forged != authorized.reference_reviews/SKILL_FORGE_RUNTIMES.md(new) — Voyager, DSPy: forge/composition vocabulary distilled; the forge mechanics are rejected as Pantheon architecture (they belong to the execution runtime).reference_reviews/SKILL_GOVERNANCE.md(new) — EviBound, SkillsVote, GovernSpec, MedSkillAudit: convergent gate/lifecycle vocabulary distilled; the gate stays a governance decision, never an autonomous approval or promotion engine.MODULES.md,AUTHORITY_INDEX.md,reference_reviews/README.md),CHANGELOG.md,ai_logs/.Inspiration, distilled not imported
The key insight: Case-Based Reasoning's retrieve/reuse/revise/retain maps almost exactly onto Pantheon's existing cap, revision and memory gates — so the loop names existing doctrine instead of adding machinery. Forge mechanics from Voyager/DSPy, dependency-graph retrieval from Graph-of-Skills, the two evidence gates from EviBound, lifecycle/contract framing from SkillsVote/GovernSpec/MedSkillAudit.
Boundary
Documentation only. No forge engine, compiler, scheduler, queue, provider router, autonomous approval engine, automatic skill installer/promotion or automatic memory promotion. Execution stays external under Task Contract. The re-evaluable professional cap (MÈTIS) and the responsibility limit remain Pantheon's own.
Governance forbidden-phrase lint reproduced locally: clean.
https://claude.ai/code/session_01UDCR8Dmp16xpFnwqJwsrPH
Generated by Claude Code