15 — TAG

Classify complete content units semantically while protecting instrument recall.

ACTIVE Built and verified

2 tool files copied and hash-verified against the original.

ML proposal-only Content graph 0 archived step 15 of 26

Where this output sits

Source evidence›Candidate evidence›Human decision›Audit acceptance›Release›Production

Source evidence, candidate evidence, human decision, audit acceptance, release and production are six separate things. This step produces the highlighted one and nothing further.

Scope

Assigns semantic categories to complete content units from a closed vocabulary, prioritising recall for instruments so that a false negative cannot quietly skip the instrument path.

This role is allowed to

Assign multiple supported labels
Route uncertainty to instrument_candidate

How it works

fed by 14 ISSUE-LINK

Inputs

Complete exact-CAP unit textROLE/TITLELayout, table/media relations and native metadata

ML proposal-only Rules/lexicon plus small supervised classifier or constrained structured AI.

No internal sequence is documented for this step — only the contract above and below it. Nothing was invented to fill the gap.

Must pass

Both instrument and instrument_candidate route to critical verification
Unknown/uncertain classes remain explicit

passes →

Outputs

Versioned multi-label class: news, advertisement, announcement, instrument, instrument_candidate, tender, table, furniture or HOLDConfidence and evidence

hands off to CRITICAL · SPELL · DGTL · ARCH

fails →

HOLD

TAG_UNKNOWNTAG_CONFLICTTAG_UNSUPPORTEDINSTRUMENT_RECALL_RISK

Nothing continues on a failed gate. Uncertainty becomes an explicit HOLD, and no later step may read an unanswered item as an accepted one.

Non-scope — what this role does NOT own

Structural roles — that is step 07
Rewriting text
Deciding release state

Explicitly forbidden

Classify isolated OCR boxes
Invent categories
Place furniture in public TOC

Dependencies and position

Starts: After complete PAGE-ORDER/ISSUE-LINK units and TITLE roles.
Previous step: 14 ISSUE-LINK
Next step: 16 CRITICAL-NOMINATE
Hands off to: CRITICAL, SPELL, DGTL, ARCH

Exact inputs

Complete exact-CAP unit text
ROLE/TITLE
Layout, table/media relations and native metadata

Exact outputs

Versioned multi-label class: news, advertisement, announcement, instrument, instrument_candidate, tender, table, furniture or HOLD
Confidence and evidence

Performer and AI/ML boundary

Performer: Rules/lexicon plus small supervised classifier or constrained structured AI.
Class: ML proposal-only
AI boundary: Allowed within a pinned, validated label set.

Training information

Not trained by this project. A pinned general model is prompted under a strict response schema. It proposes; deterministic validation accepts or HOLDs. It cannot create text or coordinates.

Deterministic validation and acceptance gates

Both instrument and instrument_candidate route to critical verification
Unknown/uncertain classes remain explicit

HOLD and failure behaviour

Failure codes this step may emit:

TAG_UNKNOWN
TAG_CONFLICT
TAG_UNSUPPORTED
INSTRUMENT_RECALL_RISK

Uncertainty becomes an explicit HOLD. Omission never converts uncertainty into acceptance, and no downstream step may treat an unanswered item as an accepted one.

Downstream handoff

CRITICAL, SPELL, DGTL, ARCH

Active tool

Status: Active tool
Run / inspect: categories applied by the MAP-AI detail call against the closed ontology set
Input: complete linked units + ROLE + exact CAP text
Output: category per group with confidence

Dependencies

Step 03 MAP-AI

Why this is the active version

ontology.json holds the closed category set actually enforced: NEWS, GOVERNMENT_INSTRUMENT, GOVERNMENT_ANNOUNCEMENT, COMPANY_ANNOUNCEMENT, TENDER, PAID_ADVERTISEMENT, TABLE, REFERENCE, HEADER, FOOTER, FURNITURE, NEEDS_REVIEW, OTHER, UNKNOWN. NEEDS_REVIEW and UNKNOWN provide the explicit uncertainty the contract requires.

Copied files — source receipt

File	Original SHA-256	Bytes	Copy
`config/ontology.json`	15b0da5dcb1df276…	1,419	verified
`config/confidence_rules.json`	534a0fd8779dbe66…	1,278	verified

2 file(s), all hash-verified against the original. Full detail in tool/SOURCE_RECEIPT.json. Copy-only: the historical source is never modified.

Archived versions

No superseded versions recorded for this role.

Known limitations

The classifier is the same constrained vision model, not a separately trained text classifier as the original design intended.
Instrument recall is not measured against an adjudicated gold set, so the recall-first requirement is asserted rather than proven.
A previous generation produced a classifier false positive and two mislabelled issues that only direct image review caught.

Security and privacy

Page images and indexed references are sent to an external provider. Credentials live outside this repository. Model output is candidate evidence, never truth.

Cost behaviour

Metered per call by the provider. Token budget and tier are set per run and recorded in the run receipts.

Copying a tool into this repository does not authorise running it, retraining it, calling a model, or processing a new issue. No paid call may be made without the owner's explicit authorisation and a hard cost cap.

Provenance

Registry: registry/roles.json (schema marsoom.e2t.roles.v1)
Derivation: 28 legacy roles - VER - HUMAN + REV = 27 active steps
Legacy role id: TAG
Legacy source SHA-256: 6f5d5d7ed8869e45307424c2193f95ffab84abc44a6a188f5c7f98c2a48ec64d
Generated: 2026-09-04T22:13:12.433Z

Feedback and decisions

Feedback is recorded per source and never merged into an invented consensus. Where sources disagree, both positions stand and the owner decides.

Source	Events
Naser / owner Final authority. Overrides every other source.	none recorded
Codex / orchestrator Architecture and sequencing.	none recorded
Builder / designer Implementation reality and constraints.	none recorded
Independent reviewer Adversarial review of claims.	none recorded
Auditor Evidence verification against artifacts.	none recorded
Human REV reviewer Page-level truth from the review site.	none recorded

Recorded events

No feedback events recorded yet. The ledger exists and is append-only: feedback/FEEDBACK_LEDGER.jsonl.

Editing feedback is not possible: a change is a new event whose supersedes names the one it replaces, and the original stays exactly as written.

Generated from registry/roles.json by tools/build-reports.mjs. Do not hand-edit — edit the registry and rebuild.
Original page pixels are the visual authority. CAP owns the exact captured text.