القائمة
Meta-fix — don't just fix the problem, fix the system that allowed the problem. Generalize fixes into patterns, policies, and prevention.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
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-metafix |
| description | Meta-fix — don't just fix the problem, fix the system that allowed the problem. Generalize fixes into patterns, policies, and prevention. |
When you fix a bug, you fixed one instance. When you meta-fix, you prevent the entire class of bugs. This skill escalates every fix through 5 levels until the root cause is addressed systemically.
/demerzel metafix [description of what just broke]
/demerzel metafix --from-pattern [pattern-id] — meta-fix a known anti-pattern
/demerzel metafix --scan — scan for fixable patterns across all state
Level 0: FIX THE INSTANCE
"The meta-grammar link is broken"
→ Fix the link
Level 1: FIX THE BATCH
"Are there other broken links?"
→ Scan and fix all broken links
Level 2: FIX THE DETECTION
"Why didn't we catch this before it was visible?"
→ Add link verification to driver RECON phase
Level 3: FIX THE PREVENTION
"Why did the link break in the first place?"
→ Add rename-tracking to the grammar prefix refactor process
Level 4: FIX THE SYSTEM
"Why is this class of problem possible at all?"
→ Create readme-sync-policy + blind-spot-detection grammar
→ Wire into driver cycle so it's automated
The meta-fix is not complete until you've reached Level 4.
What broke? Describe the specific problem.
Instance: [what happened]
Impact: [who was affected, what was wrong]
How discovered: [user noticed, audit caught, driver flagged]
Fix the immediate problem. Commit with conventional style.
Ask: "Is this the only instance, or are there others?"
Fix all siblings found. Commit.
Ask: "How should we have caught this automatically?"
Options (pick the most appropriate):
gov-blind-spot-detection.ebnfstaleness-detection-policy.yamlImplement the detection. Commit.
Ask: "Why did this happen, and how do we prevent it?"
Common root causes and their preventions:
| Root Cause | Prevention |
|---|---|
| Rename without link update | README sync policy — link check after refactors |
| New artifact without README | Coverage check — directories need READMEs |
| Stats drift from reality | Auto-count in COMPOUND phase |
| Cross-repo inconsistency | Galactic Protocol sync directives |
| Manual process skipped | Automate in driver cycle or GitHub workflow |
| Grammar gap | Research cycle grammar evolution (Step 6b) |
| No test coverage | Behavioral test for the scenario |
Implement prevention. Commit.
Ask: "What policy, grammar, or skill change ensures this entire class of problem is handled?"
Options:
state/patterns/ for future referenceImplement. Commit.
Create a meta-fix record in state/patterns/:
{
"pattern_id": "metafix-NNN-slug",
"category": "metafix",
"name": "Description of what was meta-fixed",
"levels_completed": [0, 1, 2, 3, 4],
"instance_fixed": "what broke",
"siblings_found": N,
"detection_added": "where detection was added",
"prevention_added": "what prevention was added",
"system_change": "what policy/grammar/skill was created or updated",
"timestamp": "2026-03-22T00:00:00Z"
}
Print meta-fix summary:
Meta-Fix Complete: [name]
Level 0 (instance): [what was fixed]
Level 1 (batch): [N siblings found and fixed]
Level 2 (detection): [what detection was added]
Level 3 (prevention): [what prevention was added]
Level 4 (system): [what system change was made]
Artifacts modified: [list]
Commits: [count]
/demerzel metafix --scan walks through known anti-patterns and checks:
state/patterns/anti-pattern-*.jsonLevel 0: Fix meta-grammar.ebnf → core-meta-grammar.ebnf link in org README
Level 1: Scan all READMEs — found 5 more broken links from prefix rename
Level 2: Added link verification to driver RECON phase (readme-sync-policy)
Level 3: Added rename-tracking rule: "after git mv, grep for old name in all .md files"
Level 4: Created readme-sync-policy.yaml with automated sync fields + link check
Level 0: Process 5 stale signals from 2026-03-17
Level 1: Found 30+ stale artifacts across state/
Level 2: Created staleness-detection-policy.yaml with per-category thresholds
Level 3: Added staleness scan to driver RECON phase
Level 4: Created gov-blind-spot-detection.ebnf grammar + patterns catalog
Level 0: Produce course for music department
Level 1: Found 11 of 13 departments had no courses
Level 2: Added coverage check to blind-spot-detection grammar
Level 3: Created /seldon course-pipeline for automated production
Level 4: Created grammar-evolution-policy ensuring grammars drive course production
If you find yourself repeatedly meta-fixing the same class of problem, that's a signal that the meta-fix system itself needs improvement:
This is the fixed-point question: does the meta-fix process meta-fix itself?