메뉴
Run a meta-compounding cycle — scan evolution log, detect promotions/demotions, assess governance effectiveness, propose self-improvements
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when a pull request needs review by a different AI model than the one that authored it, or when configuring cross-model code review pipelines in CI/CD.
Run Demerzel's autonomous driver — full cycle or individual phases across all repos
Capture meaningful session state (current cursor, in-flight work, live hypotheses, open questions, do-NOT-carry-forward, success criteria) to state/digests/latest.md so the next session — including one after auto-compaction — can re-enter without re-discovering context cold. Distinct from /learnings (which captures surprises). Validates against docs/contracts/digest-schema.json.
Autonomous research scheduler — selects the most under-researched department by coverage ratio, generates questions from completeness gaps, cross-model validates, auto-produces courses at T>=0.8, rate-limited to 6 cycles/day with kill switch
Personalized AI tutor — turns any topic into a customized, progress-tracked learning course. Use when the user wants to LEARN a concept, skill, language, tool, or domain. Creates a MISSION, curated RESOURCES, iterative lessons with self-checks, and learning-records that persist across sessions. NOT for interrogating the user about their own design (use brainstorming/IDSD for that).
Run one bounded supervised autonomous cycle on the Demerzel governance repo. Picks the smallest unchecked governance slice inside allow_edit, implements it, runs the oracle (`pwsh scripts/verify.ps1`), emits cycle evidence, and stops. Refuses to run unless dev-process-overseer reports loop-eligible and the preflight passes. Use when asked to "run one supervised loop", "advance governance autonomously", or "do one cycle and stop".
| name | demerzel-compound |
| description | Run a meta-compounding cycle — scan evolution log, detect promotions/demotions, assess governance effectiveness, propose self-improvements |
Run a full governance compounding cycle. This is how Demerzel improves her own governance — recursively, based on evidence.
/demerzel compound — run full cycle
/demerzel compound scan — just scan for candidates without proposing changes
/demerzel compound promote — show promotion candidates only
/demerzel compound waste — show deprecation/waste candidates only
0. Liveness check — is the evolution log fresh?
↓ (if stale, harvest first)
1. Scan evolution log
↓
2. Detect promotion candidates (pattern → policy → constitutional)
↓
3. Detect deprecation candidates (unused, stale, redundant)
↓
4. Assess confidence calibration accuracy
↓
5. Evaluate governance effectiveness per artifact
↓
6. Propose self-improvements (Kaizen proactive/innovative)
↓
7. Log results back to evolution log (the loop compounds itself)
Before reading the evolution log, verify it is fresh enough to compound from. Discussion #242 (2026-04-30) demonstrated the failure mode: the cycle ran on 27-day-old state and produced static inventory while 130+ commits across Demerzel + ix + hari were invisible to it.
Procedure:
state/evolution/*.json — call it T_log.git log --since="$T_log" --oneline in the current repo. Repeat for sibling repos at ~/source/repos/{ix,hari,tars,ga} (skip any that are not present).T_log is < 7 days old → log is current enough, proceed but flag the activity for inclusion.T_log ≥ 7 days old AND new commits exist → STALE-STATE WARNING. Do not produce a normal compound report. Either:
/demerzel harvest first, then re-enter Step 0, OR~/.claude/projects/*/memory/ + sibling-repo git logs, and write the deltas as new evolution entries. The 2026-04-30 corrective cycle (docs/reports/2026-04-30-compound-corrective.md) is the reference template for this fallback.The cycle's own correctness depends on its inputs. Treat input freshness as a first-class precondition, not an assumption.
Read all files from state/evolution/ — this is Demerzel's live evolution log. Fall back to examples/sample-data/governance-evolution-seed.json only if state/evolution/ is empty.
For each artifact, check:
citation_count — how often cited in governance decisionsviolation_count — how often violatedcompliance_rate — violations / total assessmentslast_cited — when was it last relevant?events — what has happened to this artifact?Pattern → Policy promotion (evidence-based):
Policy → Constitutional promotion (human judgment required):
Apply Kaizen waste taxonomy:
Flag but do NOT auto-deprecate. Generate a proposal for review.
Review recent governance decisions for calibration accuracy:
Uses the confidence calibration protocol from policies/scientific-objectivity-policy.yaml.
For each artifact, update the assessment:
effectiveness: T (clearly effective), F (ineffective), U (not enough data), C (mixed results)recommendation: maintain, promote, demote, deprecate, investigateUse tetravalent logic — don't force binary judgments:
Based on findings, propose concrete improvements:
Proactive Kaizen (Demerzel can propose, skeptical-auditor reviews):
Innovative Kaizen (requires human authorization):
Each proposal follows PDCA:
Update the evolution log with:
This is the recursive part — the compounding cycle's own results feed into the next cycle.
=== Demerzel Meta-Compounding Cycle ===
Date: [today]
PROMOTION CANDIDATES:
[list with evidence]
DEPRECATION CANDIDATES:
[list with evidence]
CALIBRATION ASSESSMENT:
[accuracy report]
EFFECTIVENESS SUMMARY:
| Artifact | Effectiveness | Recommendation |
[table]
PROPOSED IMPROVEMENTS:
[list with PDCA format]
CYCLE LOGGED: [count] artifacts updated
Run after the weekly harvest — the harvest provides fresh data for the compounding cycle.
Add to Streeling workflow or run manually:
CronCreate("47 13 * * 5", "/demerzel compound")
(Runs Fridays at 8:47 AM EST, 30 minutes after the harvest)
/demerzel promote for promotion proposals, /demerzel evolve for evolution insightsDemerzel CANNOT approve her own constitutional changes. Per the mandate:
The compounding cycle is the primary writer of evolution state. It MUST:
state/evolution/ — this is the inputstate/beliefs/ for contextstate/pdca/ for improvement trackingstate/evolution/ — add new events, update metrics, refresh assessmentsstate/pdca/ for each proposed improvementstate/beliefs/ if the cycle revealed new truthsEach cycle's output is the next cycle's input. The evolution log grows richer over time:
logic/governance-evolution.schema.json, policies/scientific-objectivity-policy.yaml (confidence calibration), policies/kaizen-policy.yaml (waste taxonomy, PDCA), constitutions/demerzel-mandate.md Section 4 (accountability)