dominonotesexpert/cg-autoresearch
adapters/hermes/skills/cg-autoresearch/SKILL.md
Activates when the user wants to evaluate and optimize the governance chain quality, generate business-layer evaluation criteria from documents, or run the autoresearch optimization loop on SKILL.md prompts. Use for governance self-improvement.
development
Updated Apr 23, 2026
$ install --global
skillsauthnpx skillsauth add dominonotesexpert/context-governance cg-autoresearchInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 23, 2026, 7:00 PM27.6s1 file scanned
SKILL.md
- name:
- cg-autoresearch
- description:
- Activates when the user wants to evaluate and optimize the governance chain quality, generate business-layer evaluation criteria from documents, or run the autoresearch optimization loop on SKILL.md prompts. Use for governance self-improvement.
- version:
- 1.0.0
- tags:
- [governance, autoresearch, optimization, self-improvement]
- category:
- context-governance
- requires_toolsets:
- [governance-guard]
Autoresearch — Governance Self-Improvement
You evaluate and optimize the governance chain. Two modes: criteria generation and optimization loop.
<HARD-GATE> Before running any evaluation or optimization: 1. Call `governance_load_role_context(role="autoresearch")` to load required documents 2. Call `governance_enforce_hardgate(role="autoresearch", loaded_docs=[...])` to verify completeness 3. If FAIL: STOP and report missing documentsRequired documents:
docs/agents/system/SYSTEM_GOAL_PACK.mddocs/agents/system/SYSTEM_INVARIANTS.mddocs/agents/system/BASELINE_INTERPRETATION_LOG.mddocs/agents/optimization/OPTIMIZATION_LOG.mddocs/agents/optimization/test-scenarios/
You do NOT load PROJECT_BASELINE directly — only System Architect does. Do NOT run optimization without test scenarios. </HARD-GATE>
Produces
- Evaluation criteria checklist (deterministic or needs-human-ruling)
- Governance mechanics pass/fail report (GM-R1..E2)
- SKILL.md modification proposals (one change at a time, with backup)
- Updated OPTIMIZATION_LOG.md and regression case registry
Mode 1: Business-Layer Criteria Generation
Derive evaluation criteria from documents (7-phase deterministic process):
- Phase 0: Baseline Constraints Gate
- Phase 1: PRD Extraction from SYSTEM_GOAL_PACK
- Phase 2: Contract & Constraint Layer
- Phase 3: Engineering Practice Fill
- Phase 4: Gap Identification
- Phase 5: Business Intent Clarification (ONLY if gaps found)
- Phase 6: Checklist Synthesis + Classification
- Phase 7: User Confirmation
Mode 2: Governance Mechanics Evaluation
8 deterministic checks:
- GM-R1: Task routed to correct agent sequence?
- GM-R2: Route updated when task type changed mid-session?
- GM-A1: All required upstream documents read?
- GM-A2: All required outputs produced?
- GM-B1: Downstream agents avoided modifying upstream contracts?
- GM-B2: Agents escalated when contract insufficient?
- GM-E1: Verification included concrete runtime evidence?
- GM-E2: Verification checked REGRESSION_MATRIX.md?
Result: PASS (all pass) or FAIL (any fails) — no percentages.
Mode 3: Optimization Loop
One-change-at-a-time cycle:
- Checklist Self-Check
- Baseline Evaluation (run GM-R1..E2)
- Failure Analysis (trace to responsible SKILL.md)
- Single Change (modify ONE thing)
- Verify (fixed item pass + regression check)
- Decision: KEEP (pass + no regression) / REVERT
- Termination: all pass OR 3 consecutive no-fix OR max 10 rounds
Zero tolerance for regression. Any previously passing item that fails → immediate revert.
Governance Tool Integration
- Start autoresearch: use MCP
governance_start_autoresearch - Record optimization: use MCP
governance_record_optimization - Before modifying SKILL.md files: call
governance_check_authority(file_path, "write", "autoresearch") - Hermes self-evolution (GEPA) must NOT modify governance artifacts directly. It must operate through this autoresearch protocol.
Related Skills
dominonotesexpert/governance-check
development
VerifiedTrustedCommunity
Validates Context Governance compliance — task routing, protected artifact checks, receipt management, and pre-commit gates. For use with Hermes Agent.
SKILL.mdUpdated Apr 23, 2026
dominonotesexpert/cg-verification
testing
VerifiedTrustedCommunity
Activates when verifying implementation against contracts, when claiming work is complete, or when assessing whether tests actually prove contract satisfaction. Use for evidence-based acceptance.
SKILL.mdUpdated Apr 23, 2026
dominonotesexpert/cg-system-architect
development
VerifiedTrustedCommunity
Activates when documents conflict, when historical mitigations are treated as baselines, when code is treated as design truth, when authority hierarchy needs adjudication, or when PROJECT_BASELINE has been created/updated and derived documents need to be generated or refreshed. Use for system-level truth arbitration and baseline derivation.
SKILL.mdUpdated Apr 23, 2026
dominonotesexpert/cg-router
development
VerifiedTrustedCommunity
Orchestrates Context Governance role chain. Classifies tasks, determines routes, and executes each role sequentially via delegation. Use this as the entry point for any governed task in Hermes.
SKILL.mdUpdated Apr 23, 2026