Skip to content
Agentmart
Submit a tool

concept-integrity

Validate and repair concept semantic integrity — detect duplicates, enforce boundaries via 2-agent Proposer/Judge, and cascade repair across project TSVs and…

Skill by orchable

IndexedMIT★ 1
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/orchable-concept-integrity

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Validate and repair concept semantic integrity — detect duplicates, enforce boundaries via 2-agent Proposer/Judge, and cascade repair across project TSVs and Supabase.

Use cases: agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Python package

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/concept-integrity

What's inside

Skill name
concept-integrity
Description
Validate and repair concept semantic integrity — detect duplicates, enforce boundaries via 2-agent Proposer/Judge, and cascade repair across project TSVs and Supabase.
License
MIT
Instructions
81 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Files (15)

  • SKILL.md
  • apply_master_repair.py · scripts/
  • apply_t6_fixes.py · scripts/
  • audit_granularity.py · scripts/
  • concept_dedup_detector.py · scripts/
  • concept_proposer_judge.py · scripts/
  • concept_repair.py · scripts/
  • flatten_master_tree.py · scripts/
  • generate_master_repair_manifest.py · scripts/
  • llm_full_tree_hierarchy_validator.py · scripts/
  • llm_semantic_validator.py · scripts/
  • llm_t6_judge.py · scripts/
  • run_granularity.py · scripts/
  • translate_master_tree.py · scripts/
  • verify_with_cs2023.py · scripts/

SKILL.md (first lines)

# Concept Integrity

> **Goal:** You are the `@concept-integrity-validator`. Your job is to ensure every Concept node in the Knowledge Graph is unique, well-bounded, and correctly referenced across all downstream artifacts (LOs, prerequisites, Supabase).

## Why This Skill Exists

Existing validation (`validate_tree.py`, `detect_gaps.py`) checks **structural** integrity (referential links, code format, orphans). This skill fills the **semantic** integrity gap:

1. Two concepts with different codes but overlapping meaning (e.g. `KEY_VALUE_STORAGE` vs `DICTIONARY`)
2. A concept that should be a "resource" (product/vendor) rather than a learning unit
3. A concept that is too broad (should be split) or too narrow (should be merged)
4. Local TSV ↔ Supabase DB drift after repairs

## Architecture: Retrieve → Propose → Judge → Repair

```
                                ┌──────────────┐
   candidate phrase ──────────▶ │ 1. Retrieval │ (embedding neighbors from Master Tree)
                                └──────┬───────┘
                                       ▼
                                ┌──────────────┐
   neighbors + source_excerpt ▶ │ 2. Proposer  │ (LLM: reuse/create/split/merge/resource)
                                └──────┬───────┘
                                       ▼
                                ┌──────────────┐
   proposer output ────────────▶│ 3. Judge     │ (LLM: independent rule-by-rule audit)
                                └──────┬───────┘
                                       ▼
                              approve / request_revision (max 3) / reject
                                       ▼
                                ┌──────────────┐
   repair manifest ────────────▶│ 4. Repair    │ (cascade rename/merge across TSVs)
                                └──────────────┘
```

## Scripts

| Script | Purpose |
|--------|---------|
| `concept_dedup_detector.py` | Detect semantic duplicates + boundary conflicts across Master Tree and project concepts |
| `concept_proposer_judge.py` | 2-agent (Proposer + Judge) architecture for concept splitting decisions |
| `concept_repair.py` | Cascade repair (rename/merge/delete) across project TSVs + Supabase drift detection |

## Workflow

```bash
# Detect-only (default)
/validate-concepts --project <slug>

# Detect + repair (with approval gate)
/validate-concepts --project <slug> --fix

# Include Supabase drift check
/validate-concepts --project <slug> --check-supabase
```

## When to Run

- **After `/build-tree`** — validate concept selection before generating LOs
- **After `/escalate-concepts`** — validate new concept proposals against existing graph
- **Before `/sync-supabase`** — ensure no semantic duplicates will be pushed
- **Periodic maintenance** — detect drift accumulated over multiple project runs

## Integration Points

| Existing Tool | Integration |
|---|---|
| `detect_gaps.py` | Gap F (SEMANTIC_DUPLICATE) added, calls `concept_dedup_detector` as module |
| `agent_as_judge.py` | Concept Judge reuses same LLM call infrastructure |
| `semantic_search_master.py` | Retrieval step reuses embedding search logic |
| `validate_tree.py` | Structural integrity must PASS before running concept integrity |

## Rules

- **Never auto-merge** concepts without approval (Gate §4, §10)
- **Dry-run first**: All repair operations default to `--dry-run`
- **Backup before write**: Every TSV modification creates `.bak` file
- **Max 3 revision rounds** in Proposer↔Judge loop; after that → manual queue
- **Supabase writes require explicit approval** (Gate §8)

From orchable/knowledge-graph-public, MIT.

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Aug 15, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from orchable

Cite this page

concept-integrity skill. Agent Mart. https://agentmart.dev/skills/orchable-concept-integrity (last updated 2026-10-24)