librarian-indexer

# Librarian Indexer — IPAI Edition v3.0 ## What this skill does Maintains the `/mnt/skills/user/` catalogue as a living, auditable asset: - Generates new skill stubs from a validated template - Detects drift: deprecated references, wrong Odoo version, wrong CI system - Routes skill load triggers to the correct skill bundles - Writes or updates `skills-registry.json` entries - Produces Azure Pipelines lint gates for skill hygiene **CI/CD doctrine**: Azure Pipelines is the sole CI/CD authority. GitHub Actions is FORBIDDEN — remove any reference on sight. --- ## §1 — IPAI skill inventory (canonical) ### `/mnt/skills/user/` — current catalogue | Skill folder | Domain | Load trigger | |---|---|---| | `ipai-platform` | Master bundle — Azure resources, agent routing, WAF | "IPAI", "InsightPulseAI", "Pulser", any Azure resource name | | `ipai-odoo-platform` | Odoo 18 CE dev, OCA-first, BIR compliance, ACA deploy | "Odoo", "OCA", "ipai_*", "BIR", "`<list>`", "view_mode" | | `ipai-resource-map` | Azure resource constants — MIs, RGs, sub IDs | Any IPAI Azure resource lookup | | `ipai-agent-platform` | MAF patterns, Foundry Agent Service, judge infra | "agent", "MAF", "Foundry agent", "judge" | | `odoo18-ce-development` | Odoo 18 CE Smart Delta, OCA conventions, ACA deploy | "Odoo 18", "delta module", "Smart Delta" | | `prisma2020-flow-diagram` | PRISMA 2020 R + Python flow diagrams | "PRISMA", "systematic review", "flow diagram" | | `doing-meta-analysis-r` | Harrer et al. R meta-analysis methodology | "meta-analysis", "forest plot", "heterogeneity", "I²" | | `meta-analysis-pipeline` | Python PRISMA + forest/funnel without R | "Python meta-analysis", "pub bias", "PRISMA Python" | | `consulting-writeup-pwc` | PwC-style reports, SCQA, answer-first | "consulting", "PwC", "board-ready", "executive summary" | | `editorial-pipeline-tools` | Schema-first linting, YAML-to-deck CI | "editorial CI", "linter", "SCQA validate" | | `editorial-print-production` | Print-ready PDF, InDesign handoff, CMYK | "print", "InDesign", "CMYK", "bleed" | | `design-system-engineer` | Fluent 2, Material 3, Fiori token translation | "design system", "Fluent", "token", "component" | | `figma-design-to-code` | Figma MCP → React/Tailwind/Odoo code | "Figma", "design-to-code", "Code Connect" | | `fluent-perf-testing` | Playwright + Lighthouse CI for Fluent 2 UI | "performance test", "Lighthouse", "Fluent perf" | | `mcp-complete-guide` | Production MCP server, 11-phase guide | "MCP server", "tool schema", "semantic layer" | | `drawio-diagrams-enhanced` | draw.io XML + PMBOK shapes | "draw.io", "BPMN", "swimlane", "WBS" | | `pmbok-project-management` | PMBOK 7, risk, schedule, cost control | "PMBOK", "project management", "risk register" | | `librarian-indexer` | This skill — catalogue management | "skill index", "catalogue", "skill audit", "librarian" | ### `/mnt/skills/public/` — platform skills (read-only) `docx`, `pdf`, `pdf-reading`, `pptx`, `xlsx`, `frontend-design`, `file-reading`, `product-self-knowledge` ### `/mnt/skills/examples/` — example skills (read-only) `web-artifacts-builder`, `skill-creator`, `mcp-builder`, `brand-guidelines`, `theme-factory` --- ## §2 — Skill metadata schema Every skill MUST have a valid YAML frontmatter block: ```yaml --- name: <kebab-case-name> description: > One-sentence trigger description (what Claude should load this skill for). Include: domain keywords, primary use cases, what it replaces/pairs with. ≤ 200 characters for accurate trigger matching. --- ``` ### Extended registry entry (skills-registry.json) ```jsonc { "name": "ipai-bir-tax-compliance", "path": "/mnt/skills/user/ipai-bir-tax-compliance/SKILL.md", "domain": "finance", "plane": "CoreOps", "ado_area": "InsightPulseAI\\CoreOps", "odoo_version": "18.0", "depends_on": ["ipai-odoo-platform", "ipai-resource-map"], "agents": ["tax_guru"], "deprecated_refs": [], "last_audited": "2026-04-15", "status": "active" } ``` --- ## §3 — Drift detection rules Run this audit before publishing any skill update. Flag as errors: | Pattern | Rule | Action | |---|---|---| | `GitHub Actions` / `.github/workflows` | FORBIDDEN — Azure Pipelines only | Remove entirely | | `Supabase` / `supabase` | DEPRECATED | Remove; replace with `pg-ipai-odoo platform schema` | | `Vercel` / `vercel` | DEPRECATED | Remove; replace with ACA | | `n8n` / `n8n.io` | DEPRECATED | Remove; replace with Service Bus | | `DigitalOcean` / `DO droplet` | DEPRECATED | Remove | | `Cloudflare` / `cloudflare` | DEPRECATED | Remove | | `nginx` | DEPRECATED | Remove | | `Odoo 19` / `odoo19` / `19.0.*` in version | Wrong version | Replace with Odoo 18 CE / `18.0.*` | | `<tree>` in Odoo view XML | Deprecated tag | Replace with `<list>` | | `view_mode="tree` | Deprecated value | Replace with `view_mode="list,form"` | | `DockerHub` / `docker.io` | Not IPAI registry | Replace with `acripaidev.azurecr.io` | | `secrets.GITHUB_TOKEN` | GitHub Actions secret | Remove (Azure Pipelines uses service connection) | | `Supabase project: spdtwktxdalcfigzeqrz` | Stale credential hint | Remove | | `pgvector` without `pg-ipai-odoo` context | Missing resource context | Add `pg-ipai-odoo.postgres.database.azure.com` | ### Drift scan (bash) ```bash # Run from repo root — exits non-zero if violations found grep -rn \ "GitHub Actions\|\.github/workflows\|supabase\|Supabase\|Vercel\|vercel\|n8n\|DigitalOcean\|cloudflare\|Cloudflare\|DockerHub\|dockerhub\|Odoo 19\|odoo19\|<tree>\|view_mode=\"tree\|nginx" \ /mnt/skills/user/ \ --include="*.md" \ | grep -v "DEPRECATED\|deprecated\|removed\|do not use" \ | tee /tmp/skill-drift.txt VIOLATIONS=$(wc -l < /tmp/skill-drift.txt) echo "Drift violations: $VIOLATIONS" [ "$VIOLATIONS" -eq 0 ] && echo "PASS" || exit 1 ``` --- ## §4 — Skill generation template Use this template when creating a new `ipai_*` skill from scratch: ```markdown --- name: <skill-name> description: > <Trigger description — what Claude loads this for. Include domain keywords.> --- # <Skill title> — IPAI Edition ## Purpose <One paragraph. What problem does this solve? Which agents/modules does it serve?> ## Scope - Odoo version: 18.0 CE - OCA-first: yes — list OCA modules this depends on - ipai_* delta: yes/no — only when CE+OCA cannot cover - Azure resources: list canonical names (ipai-copilot-resource, pg-ipai-odoo, etc.) - Agents: list provisioned agents this skill serves ## Prerequisites - [ ] `ipai-odoo-platform` loaded (for Odoo 18 CE conventions) - [ ] `ipai-resource-map` loaded (for Azure resource names) ## Core patterns ### [Pattern name] <Code, config, or structured guidance. Production-grade only.> ## Odoo conventions (enforced) - Views: `<list>` always — never `<tree>` - `view_mode="list,form"` — never `"tree,form"` - Module version: `18.0.x.x.x` - Depends: OCA modules first, `ipai_*` delta last ## Azure conventions (enforced) - CI/CD: Azure Pipelines only (`azure-pipelines/`) - GitHub Actions: FORBIDDEN - Secrets: `kv-ipai-dev-sea` — never in code - Registry: `acripaidev.azurecr.io` — never DockerHub - Identity: `DefaultAzureCredential` → `ManagedIdentityCredential` in prod ## Artifact paths | Output | Target path | |---|---| | Odoo module | `odoo/addons/ipai/<module_name>/` | | Azure IaC | `infra/azure/<resource-type>/*.bicep` | | Pipeline | `azure-pipelines/*.yml` | | Agent skill | `agents/skills/<domain>/SKILL.md` | | Platform contract | `platform/contracts/*.yaml` | | Runbook | `docs/runbooks/*.md` | ## Related skills - Pairs with: `ipai-odoo-platform`, `ipai-resource-map` - ADO area path: `InsightPulseAI\<CoreOps|AdsIntel|Research|DataIntel|Platform|ReleaseOps>` ``` --- ## §5 — Skill trigger routing (IPAI-specific) Load triggers for Claude — map user intent to correct skill bundles: ```yaml triggers: # Core operations odoo_general: keywords: ["Odoo", "OCA", "module", "ipai_*", "addons", "manifest"] load: [ipai-odoo-platform, ipai-resource-map] bir_tax: keywords: ["BIR", "2307", "2550M", "1601-C", "SAWT", "ATC code", "eBIRForms"] load: [ipai-odoo-platform, ipai-resource-map] agent: tax_guru bank_recon: keywords: ["reconciliation", "account_reconcile", "bank statement", "period close"] load: [ipai-odoo-platform, ipai-resource-map] agent: bank_recon finance_close: keywords: ["month-end", "mis_builder", "D+5", "period lock", "Finance Close"] load: [ipai-odoo-platform, ipai-resource-map] agent: finance_close # Agent platform agent_build: keywords: ["MAF", "agent framework", "Foundry agent", "skill file", "judge"] load: [ipai-agent-platform, ipai-resource-map] mcp_server: keywords: ["MCP server", "tool schema", "ipai-odoo-mcp", "FastMCP", "StreamableHTTP"] load: [mcp-complete-guide, ipai-agent-platform] # Azure / infra azure_infra: keywords: ["Bicep", "ACA", "pg-ipai-odoo", "kv-ipai-dev-sea", "rg-ipai"] load: [ipai-resource-map, ipai-platform] ci_cd: keywords: ["azure-pipelines", "pipeline", "build agent", "ipai-build-agent"] load: [ipai-resource-map] note: "Azure Pipelines only — never GitHub Actions" # Research / PrismaLab systematic_review: keywords: ["PRISMA", "systematic review", "meta-analysis", "forest plot", "RoB"] load: [prisma2020-flow-diagram, doing-meta-analysis-r, meta-analysis-pipeline] # Creative / media w9_studio: keywords: ["W9 Studio", "TVC", "content job", "booking", "Sora 2", "finishing"] load: [ipai-resource-map] # Skill management skill_audit: keywords: ["skill", "catalogue", "index", "drift", "librarian"] load: [librarian-indexer] ``` --- ## §6 — Azure Pipelines gate for skill hygiene Add to `azure-pipelines/skill-audit.yml`: ```yaml # azure-pipelines/skill-audit.yml trigger: paths: include: [agents/skills/**/*.md, docs/**/*.md] pool: name: ipai-build-agent jobs: - job: SkillDriftCheck displayName: Skill drift audit steps: - script: | VIOLATIONS=$(grep -rn \ "GitHub Actions\|\.github/workflows\|supabase\|Supabase\|Vercel\|n8n\|DigitalOcean\|Cloudflare\|DockerHub\|Odoo 19\|<tree>\|view_mode=\"tree" \ agents/skills/ docs/ \ --include="*.md" \ | grep -v "DEPRECATED\|deprecated\|removed" \ | wc -l) echo "Drift violations: $VIOLATIONS" [ "$VIOLATIONS" -eq 0 ] && echo "##vso[task.setvariable variable=driftClean]true" || \