meta

Meta Skill

meta is the governance control-plane orchestrator for this skill corpus. It is responsible for recursive decomposition, deterministic artifact generation, quality gating, and release packaging.

Completion Promise (Initial User Prompt)

This skill explicitly fulfills the original request to:

1. Classify and interconnect all provided skills through textual concept extraction and graph analytics.
2. Decompose and reorganize them into a modular, cluster-first structure with symlink compatibility.
3. Emit a complete control-plane artifact graph (.index/*) with explicit hard/soft dependency semantics, bridge metadata, and ontology.
4. Optimize recursively to a DAG-friendly topology using cycle demotion and transitive-reduction while preserving compatibility.
5. Package and validate the result as a reusable skill bundle and plugin-ready skill package under strict checks (MCTSR >= 95, hard DAG acyclicity, no unresolved references, clean release manifest).

Fast Path

python .index/interfaces/emit_index.py --base .

Full Deterministic Loop (recommended)

python skills/meta/scripts/meta_cycle.py --base . --iterations 5 --strict --package

The loop emits .index artifacts, regenerates a BD-style issue ledger, and (when requested) creates meta_release.zip.

Progressive Loading Model

1. Resolve compatibility aliases and manifest metadata.
2. Compute raw and reduced hard dependency graphs.
3. Build reverse dependencies and SCC condensation.
4. Promote uncovered couplings into soft_refs when necessary.
5. Emit ontology/control payloads and quality reports.
6. Record unresolveds and action items as an issue ledger.
7. Package release artifact set with checksums.

RALPH Orchestration

meta runs a Recurrentive, Analytic, Learning-guided, Pruning/hardening loop:

1. Request replay: read the latest graph, manifests, and constraints.
2. Analysis pass: emit raw graph artifacts (skill_graph, clustered_graph, ontology schema payloads).
3. Leveling: apply SCC decomposition, reverse-call synthesis, and bridge/soft-edge augmentation.
4. Pruning: demote cycle edges to soft refs and apply transitive reduction on DAG edges.
5. Hardening: generate BD ledger, run strict validation, and package release artifacts.

Use this explicit form with:

python skills/meta/scripts/meta_cycle.py --base . --iterations 5 --strict --package

Runtime dependencies

• Hard dependency: system-skill, skill-orchestrator
• Soft compatibility references: skill-protocol, skill-updater

Issue Tracking (BD loop)

• Generate issue ledger:

python skills/meta/scripts/issue_ledger.py --base . --output .index/bd_ledger.json --strict

• Interpret severity tiers from severity (critical, high, medium, low, info) and address all critical findings before release.

Packaging (no legacy bundle artifacts)

python skills/meta/scripts/package_release.py --base . --strict

Outputs ./.index/meta_release.zip and ./.index/meta_release_manifest.json.

Inputs and outputs

Inputs

• skills/*/manifest.json
• skills/*/bridge.json
• repo/**/.skill
• README.md

Outputs

• .index/main_index.yaml
• .index/main_index_payload.json
• .index/quality_report.md
• .index/processing_report.md
• .index/bridge_candidates.csv
• .index/bridge_index.json
• .index/clustered_graph.json
• .index/skill_graph.json
• .index/control_ontology.json
• .index/control_ontology.schema.json
• .index/hyperedges.json
• .index/issue_log.json
• .index/bd_ledger.json
• .index/meta_release.zip
• .index/meta_release_manifest.json

Multi-platform integration

• agents/claude.json: direct command for Claude Code runtime entry.
• agents/codex.json: direct command for Codex runtime entry.
• agents/gemini.json: direct command for Gemini CLI integration.
• agents/agent-skills.json: Agent Skills platform compatibility.
• agents/openai.yaml: OpenAI Agent interface compatibility.
• Repository-level meta package metadata is mirrored by .claude-plugin/marketplace.json.

Notes

• This skill is deterministic: repeated runs over the same graph should converge and show stable metrics.
• If unresolved references appear, rebuild manifests first, then re-run the loop.