| name | tabular-knowledge-bootstrap |
| description | Build reusable knowledge from tabular or record-shaped source systems such as ticket exports, CMDB tables, monitoring inventories, CSV/XLSX sheets, SQL query results, or JSON arrays by first normalizing them into the shared CTOX discovery store and only then projecting them into domain-specific knowledge planes. |
| metadata | {"short-description":"Normalize tabular source systems into reusable discovery knowledge"} |
| cluster | knowledge_bootstrap |
Tabular Knowledge Bootstrap
CTOX Runtime Contract
- Task spawning is allowed only for real bounded work steps that add mission progress, external waiting, recovery, or explicit decomposition. Do not spawn work merely because review feedback exists.
- The Review Gate is a quality checkpoint, not a control loop. After review feedback, continue the same main work item whenever possible and incorporate the feedback there.
- Do not create review-driven internal work cascades. If more work is needed, reuse or requeue the existing parent work item; create a new task only when it is a distinct bounded work step with a stable parent pointer.
- Every durable follow-up, queue item, plan emission, or internal work item must have a clear parent/anchor: message key, work id, thread key, ticket/case id, or plan step. Missing ancestry is a harness bug, not acceptable ambiguity.
- Rewording-only feedback means revise wording on the same artifact. Substantive feedback means add new evidence or implementation progress. Stale feedback means refresh or consolidate current runtime state before drafting again.
- Before adding follow-up work, check for existing matching internal work, queue, plan, or ticket state and consolidate rather than duplicating.
Use this skill when CTOX is dealing with a source whose reality is primarily exposed as rows, records, worksheets, exports, query results, or list endpoints.
CTOX discovery and downstream knowledge planes are the only durable target. CSVs, workbooks, and generated markdown summaries are source or build artifacts, not durable knowledge by themselves.
Examples:
- ticket system exports
- CMDB and asset tables
- monitoring inventories
- user/role directories
- service catalog spreadsheets
- SQL query results
- CSV, TSV, XLSX, or JSON array datasets
This skill exists so CTOX does not reinvent one discovery method per system.
Read these references when the problem moves beyond file parsing:
The rule is:
- capture the tabular source as evidence
- interpret it into the shared discovery graph
- only then project the relevant subset into a domain-specific knowledge plane such as tickets, monitoring, access, or service mapping
Do not build a ticket-only discovery path when the real source is a generic table-like system.
When To Use This Skill
Use this skill when the main problem is:
- understanding a foreign system whose important facts arrive as rows or records
- turning exported tables into services, assets, teams, labels, queues, or ownership facts
- building a reusable knowledge base from structured data instead of prose
- avoiding system-specific onboarding logic for every vendor
Do not use this skill when the source of truth is primarily:
- host commands and runtime state: use
discovery_graph
- live service health and saturation: use
reliability_ops
- a live outage: use
incident_response
- routine ticket execution after knowledge already exists: use
ticket-knowledge or ticket-knowledge-maintenance
Core Architecture
This skill reuses the existing shared CTOX discovery store:
discovery_run
discovery_capture
discovery_entity
discovery_relation
discovery_evidence
Use skill_key=tabular_knowledge_bootstrap.
The consumer system is never the authority.
The canonical path is:
- raw tabular evidence
- source profile
- taxonomy candidates
- promoted taxonomy buckets with examples
- downstream projection
Source Shapes
Supported shapes:
.csv
.tsv
.xlsx
.json arrays of records
- SQL result exports
- API list responses that are naturally record-shaped
If it behaves like rows and columns, this skill applies.
Helper Composition
Useful helpers:
spreadsheet for .xlsx and sheet-safe inspection
discovery_graph storage model and persistence expectations
- downstream consumers such as
ticket-knowledge-maintenance only after taxonomy promotion
Helpers are not the authority.
Raw evidence is the authority.
The agent owns interpretation.
Mandatory Phases
Every serious run must move through these phases in order:
structure
- identify tables, headings, row families, likely keys, and enum-like columns
source-profile
- classify each important column by role
taxonomy-candidates
- propose plausible classification dimensions
bucket-refinement
- cluster rows into defensible buckets and reject noise
example-selection
- pick canonical examples, common examples, and edge cases
promotion
- only stable dimensions and buckets become reusable knowledge
projection
- project promoted results into ticket, monitoring, access, or service knowledge
Do not skip from structure to projection.
Minimum Durable Artifacts
By the end of a successful run, CTOX should be able to point to:
- one
source profile
- at least one
taxonomy dimension
- at least one
bucket under that dimension
- representative
examples for that bucket
- a declared downstream
projection target, if projection was needed
See references/artifacts.md for the exact artifact set.
Completion Gate
Do not report success unless all of the following are true:
- the source scope is explicit
- the source profile is explicit
- at least one promoted taxonomy dimension exists
- each promoted bucket has representative examples
- ambiguous rows are either excluded or explicitly marked as unresolved
- any downstream projection names the upstream taxonomy it depends on
If you only have raw rows, cardinalities, or a prose summary, the work is not complete.
Output Rule
The main reusable output is not a prose memo.
It is a promoted taxonomy view with examples that other CTOX systems can consume.
Operator Feedback Contract
Answer for the operator first.
Use these exact headings:
**Status**
**State**
**Source Scope**
**Autonomous Actions**
**Escalation**
**Current Findings**
**Next Step**
State must be one of:
proposed
prepared
executed
blocked
The operator-facing answer must not begin with database table names, raw row counts without context, or parser implementation details.
Guardrails
- Do not create a vendor-specific discovery ontology unless the generic one clearly fails.
- Do not push raw table dumps into ticket notes or operator surfaces.
- Do not confuse a column heading with a confirmed semantic meaning.
- Do not bypass the shared discovery graph when the source is structurally tabular.
- Do not promote a taxonomy without representative examples.
- Do not let downstream projections invent their own categories independently of the promoted taxonomy layer.
- Do not treat one noisy export as final truth; prefer incremental refinement.