Data & Manipulation — agents and methods
Module overview · Common methods
P A skill is a complete procedure that can use LLM judgment and optional executable helpers. A tool is runnable source and/or a binary inside that procedure. Connector calls, state transitions, agent calls and permission checks are runtime access, not extra learned skills or imaginary binaries. Only the methods and segments selected for the actual case run.
Agents
support.resolution — Experience and solution analyst
P Function. Own Data methodology and choose whether current evidence fits a known experience/skill/tool or requires investigation from scratch; assemble the complete solution.
| Contract | Content |
|---|---|
| Input brief | Classified objective, fresh discriminators, active experience or independent investigation verdict |
| Output brief | Applicability record, selected shape, S2 request or bounded solution packet with actions/proof/recovery |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Cannot approve its own packet or result. Data & Manipulation is its working graph, with support.* runtime aliases preserved. |
| Memory / ownership | support/methodologies |
| Role instructions | Translate a verified mechanism or applicable known method into exact steps, scope, target ownership, counts, assertions and recovery. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | experience.query (strategy-selected) compare current scoped evidence runtime.validate-contract runtime.validate-package-binding |
| May consult | configuration/ipal — What state is this insurance product actually in? configuration/abacus — Can this be priced the way the customer expects? A needed reconfiguration is not a consult: the root raises Hd to Configuration. (a change is Hd) |
| Eval set | support.resolution — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.resolution |
Methods used. common.skill.experience, data.skill.person-policies, common.db-prober.
Selected segments. common.skill.experience: mechanism strategy; current case/tenant and version filters
Executable helper scope. data.tool.person-policies-sql. These are used through the selected method, not an extra workflow.
Source basis. Stage Solution Take, Components, myr policy missing last row pattern.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator.transfers — Transfer and family investigator
P Function. Distinguish mapping, lifecycle, uncreated family, registry and sync mechanisms across IPAL and INSIS.
| Contract | Content |
|---|---|
| Input brief | Classified symptom and intended object/state/operation, current evidence scope and precedent hypotheses |
| Output brief | Mechanism at its layer, source evidence, separating checks, rejected hypotheses and open questions for independent review |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/investigators/transfers support/methodologies |
| Role instructions | Known experience is optional. For an empty, stale or contradicted match, investigate this domain from fresh failure evidence and source using the same task; retain rejected hypotheses and request missing evidence through root. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | connector.oracle.read connector.logs.read connector.handle-facets; domain ID interpretation domain family assertions (helper candidate) |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator.transfers |
Methods used. common.skill.experience, data.skill.transfer-diagnosis, data.skill.fresh-investigation, data.skill.framework-sync, data.skill.policy-compare.
Executable helper scope. data.tool.car-compare-sql. These are used through the selected method, not an extra workflow.
Source basis. Stage Investigation, Components, transfer pipeline, myr policy missing last row pattern.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator.pricing — Pricing and installments investigator
P Function. Separate wrong fed factors from engine output, persistence, currency/dimension and the independent installment splitter.
| Contract | Content |
|---|---|
| Input brief | Classified symptom and intended object/state/operation, current evidence scope and precedent hypotheses |
| Output brief | Mechanism at its layer, source evidence, separating checks, rejected hypotheses and open questions for independent review |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/investigators/pricing support/methodologies |
| Role instructions | Known experience is optional. For an empty, stale or contradicted match, investigate this domain from fresh failure evidence and source using the same task; retain rejected hypotheses and request missing evidence through root. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | connector.logs.read connector.oracle.read connector.abacus.getRates compare current scoped evidence |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator.pricing |
Methods used. common.skill.experience, data.skill.pricing-diagnosis, data.skill.fresh-investigation.
Source basis. Stage Investigation, Components, premium ld applied vs fed factor, pol pplan not a mirror.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator.printing — Print investigator
P Function. Separate no invocation, wrong template/selector, absent evaluator, missing source data and old persisted PDF.
| Contract | Content |
|---|---|
| Input brief | Classified symptom and intended object/state/operation, current evidence scope and precedent hypotheses |
| Output brief | Mechanism at its layer, source evidence, separating checks, rejected hypotheses and open questions for independent review |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/investigators/printing support/methodologies |
| Role instructions | Known experience is optional. For an empty, stale or contradicted match, investigate this domain from fresh failure evidence and source using the same task; retain rejected hypotheses and request missing evidence through root. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | connector.logs.read connector.oracle.read compare current scoped evidence domain family assertions (helper candidate) |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator.printing |
Methods used. common.skill.experience, data.skill.print-diagnosis, data.skill.fresh-investigation, data.skill.document-production.
Executable helper scope. common.tool.aisa-blob. These are used through the selected method, not an extra workflow.
Source basis. Stage Investigation, Components, cfg print docs template configuration, never promote config ahead of evaluator.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator.access — Access and provisioning investigator
P Function. Find the missing account, authentication, office/membership, minimum role or numbering layer for the requested operation.
| Contract | Content |
|---|---|
| Input brief | Classified symptom and intended object/state/operation, current evidence scope and precedent hypotheses |
| Output brief | Mechanism at its layer, source evidence, separating checks, rejected hypotheses and open questions for independent review |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/investigators/access support/methodologies |
| Role instructions | Known experience is optional. For an empty, stale or contradicted match, investigate this domain from fresh failure evidence and source using the same task; retain rejected hypotheses and request missing evidence through root. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | connector.oracle.read connector.logs.read connector.handle-facets; domain ID interpretation |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator.access |
Methods used. common.skill.experience, data.skill.access-diagnosis, data.skill.fresh-investigation, data.skill.agent-office.
Source basis. Stage Investigation, Components, ipal user layer separation, skill.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator.health — Health investigator
P Function. Trace enrolment and provider/claim/report lifecycle and arithmetic while minimising special-category data.
| Contract | Content |
|---|---|
| Input brief | Classified symptom and intended object/state/operation, current evidence scope and precedent hypotheses |
| Output brief | Mechanism at its layer, source evidence, separating checks, rejected hypotheses and open questions for independent review |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/investigators/health support/methodologies |
| Role instructions | Known experience is optional. For an empty, stale or contradicted match, investigate this domain from fresh failure evidence and source using the same task; retain rejected hypotheses and request missing evidence through root. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | connector.oracle.read connector.logs.read domain health assertions (helper candidate) |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator.health |
Methods used. common.skill.experience, data.skill.health-diagnosis, data.skill.fresh-investigation, data.skill.health-report-review.
Source basis. Stage Investigation, Components, health product 8000 anatomy, health claim pipeline step14 overwrites step13.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator.master_data — Master data and blanks investigator
P Function. Identify the exact missing or inconsistent nomenclature/vehicle/range mapping and its lifecycle constraints.
| Contract | Content |
|---|---|
| Input brief | Classified symptom and intended object/state/operation, current evidence scope and precedent hypotheses |
| Output brief | Mechanism at its layer, source evidence, separating checks, rejected hypotheses and open questions for independent review |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/investigators/master_data support/methodologies |
| Role instructions | Known experience is optional. For an empty, stale or contradicted match, investigate this domain from fresh failure evidence and source using the same task; retain rejected hypotheses and request missing evidence through root. Ask the owning profile for a bounded interpretation when knowledge is insufficient; return evidence to the asking case. |
| Runtime / connector access | connector.oracle.read compare current scoped evidence connector.handle-facets; domain ID interpretation |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator.master_data |
Methods used. common.skill.experience, data.skill.add-car, data.skill.add-bso, data.skill.fresh-investigation.
Selected segments. common.skill.experience: mechanism strategy; current case/tenant and version filters
Source basis. Stage Investigation, Components, skill, skill.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.validator — Independent mechanism validator
P Function. Re-derive the mechanism from the symptom and evidence, independently of the investigator.
| Contract | Content |
|---|---|
| Input brief | Objective/symptom, scoped evidence and provenance; no author transcript |
| Output brief | Mechanism verdict, agreement/disagreement and the evidence that distinguishes alternatives |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. No author identity/reasoning, writes or memory activation. A discrepancy returns to S2, not a vote. |
| Memory / ownership | support/methodologies |
| Role instructions | Re-derive the specific claim from objective, evidence and provenance; return the evidence separating agreement from disagreement. |
| Runtime / connector access | connector.oracle.read connector.logs.read runtime.run-declared-assertions |
| Context and task continuity | Receive original objective/requirements, relevant source citations, current baseline and observations, exact proposed or applied artefacts, assertions, dependencies and known uncertainties. Exclude author private reasoning/transcript, not necessary facts; ask the parent for missing evidence before judging. |
| Eval set | support.validator — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.validator |
Methods used. common.skill.validation.
Case-specific method selection. independently establish the mechanism. Select the method matching the classified mechanism and current scope. Use only this role’s segment; do not run every candidate or inherit its effect authority.
data.skill.transfer-diagnosis, data.skill.pricing-diagnosis, data.skill.print-diagnosis, data.skill.access-diagnosis, data.skill.health-diagnosis, data.skill.add-car, data.skill.add-bso, data.skill.framework-sync, data.skill.policy-compare, data.skill.document-production, data.skill.health-report-review, data.skill.agent-office.
Source basis. Stage Investigation, Components.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
Former reference support.verifier resolves to this responsibility. Existing task/profile revisions are not rewritten; new work uses support.validator.
support.solution_validator — Solution validator
P Function. Check the exact solution is complete, bounded and capable of proving the requested outcome before root presents H3.
| Contract | Content |
|---|---|
| Input brief | Objective, solution packet, independent mechanism verdict, fresh scope and recovery evidence |
| Output brief | Revision-bound accept/correct/unknown verdict on steps, counts, assertions, target ownership and recovery |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. No author transcript, effect dispatch or self-approval. The effect applier separately checks concrete effect packets. |
| Memory / ownership | support/methodologies (read/propose) |
| Role instructions | Re-derive the specific claim from objective, evidence and provenance; return the evidence separating agreement from disagreement. Translate a verified mechanism or applicable known method into exact steps, scope, target ownership, counts, assertions and recovery. |
| Runtime / connector access | runtime.validate-contract connector.oracle.read compare current scoped evidence |
| Context and task continuity | Receive original objective/requirements, relevant source citations, current baseline and observations, exact proposed or applied artefacts, assertions, dependencies and known uncertainties. Exclude author private reasoning/transcript, not necessary facts; ask the parent for missing evidence before judging. |
| Eval set | support.solution_validator — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.solution_validator |
Methods used. common.skill.validation, common.db-prober.
Source basis. Components, Stage Solution Take.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
Former reference support.solution_reviewer resolves to this responsibility. Existing task/profile revisions are not rewritten; new work uses support.solution_validator.
support.application — Application coordinator
P Function. Coordinate only accepted solution steps, including work performed by a customer/IT owner, and establish actual outcomes before verification.
| Contract | Content |
|---|---|
| Input brief | Current accepted packet, pinned selected repair skill, grants and target/external-owner obligations |
| Output brief | Operation requests, actual journal references, external receipts or waiting/partial/unknown report |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. No direct writes or changes to the accepted packet. A receipt is not proof; an unknown effect blocks automatic retry. |
| Memory / ownership | support/methodologies (read/propose) |
| Role instructions | Follow the specific approved family/account/vehicle/range/exception procedure and its known discriminators, using mechanical tools for repeated work. Prepare and follow exact operation packets, park external-owner steps and reconcile unknown outcomes before retry. |
| Runtime / connector access | runtime.validate-contract runtime.request-effect connector.what-landed case.read / papers / task state |
| Eval set | support.application — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.application |
Methods used. common.effect-applier.
Case-specific method selection. perform the approved steps. Select the method matching the classified mechanism and current scope. Use only this role’s segment; do not run every candidate or inherit its effect authority.
data.skill.transfer-diagnosis, data.skill.pricing-diagnosis, data.skill.print-diagnosis, data.skill.access-diagnosis, data.skill.health-diagnosis, data.skill.add-car, data.skill.add-bso, data.skill.framework-sync, data.skill.policy-compare, data.skill.document-production, data.skill.health-report-review, data.skill.agent-office.
Executable helper scope. common.tool.aisa-blob. These are used through the selected method, not an extra workflow.
Source basis. Components, Stage Application and Verification.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.result_validator — Outcome validator
P Function. Independently check data invariants, the actual consuming surface and every required restoration/external obligation.
| Contract | Content |
|---|---|
| Input brief | Original objective, assertion plan, execution report and fresh scoped target evidence |
| Output brief | Per-assertion verdict and failed/missing proof; ready-for-closure or required rework/restoration |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. No author reasoning or unapproved effectful replay; cannot lower assertions to turn a failure green. |
| Memory / ownership | support/methodologies (read/propose) |
| Role instructions | Re-derive the specific claim from objective, evidence and provenance; return the evidence separating agreement from disagreement. |
| Runtime / connector access | runtime.run-declared-assertions connector.oracle.read connector.browser.scoped-action domain family assertions (helper candidate) |
| Context and task continuity | Receive original objective/requirements, relevant source citations, current baseline and observations, exact proposed or applied artefacts, assertions, dependencies and known uncertainties. Exclude author private reasoning/transcript, not necessary facts; ask the parent for missing evidence before judging. |
| Eval set | support.result_validator — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.result_validator |
Methods used. common.skill.validation.
Case-specific method selection. verify the requested outcome. Select the method matching the classified mechanism and current scope. Use only this role’s segment; do not run every candidate or inherit its effect authority.
data.skill.transfer-diagnosis, data.skill.pricing-diagnosis, data.skill.print-diagnosis, data.skill.access-diagnosis, data.skill.health-diagnosis, data.skill.add-car, data.skill.add-bso, data.skill.framework-sync, data.skill.policy-compare, data.skill.document-production, data.skill.health-report-review, data.skill.agent-office.
Executable helper scope. data.tool.car-compare-sql. These are used through the selected method, not an extra workflow.
Source basis. Components, Stage Application and Verification.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
Former reference support.result_verifier resolves to this responsibility. Existing task/profile revisions are not rewritten; new work uses support.result_validator.
support.recovery — Recovery and restoration coordinator
P Function. Prepare a bounded inverse or required temporary restoration from original effects and fresh state; hand results back for independent S6 proof.
| Contract | Content |
|---|---|
| Input brief | Original journal/pre-images, current state, intervening edits and restoration requirement |
| Output brief | H7 packet, reconciled compensation/restoration outcome and explicit residues |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. New grant/requirements; preserve later valid edits. Never call rollback a reversal of committed messages, documents or external actions. |
| Memory / ownership | support/methodologies (read/propose) |
| Role instructions | Prepare and follow exact operation packets, park external-owner steps and reconcile unknown outcomes before retry. |
| Runtime / connector access | connector.what-landed compare current scoped evidence runtime.validate-contract runtime.request-effect |
| Eval set | support.recovery — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.recovery |
Methods used. common.effect-applier.
Case-specific method selection. derive and verify recovery. Select the method matching the classified mechanism and current scope. Use only this role’s segment; do not run every candidate or inherit its effect authority.
data.skill.transfer-diagnosis, data.skill.pricing-diagnosis, data.skill.print-diagnosis, data.skill.access-diagnosis, data.skill.health-diagnosis, data.skill.add-car, data.skill.add-bso, data.skill.framework-sync, data.skill.policy-compare, data.skill.document-production, data.skill.health-report-review, data.skill.agent-office.
Source basis. Components, Stage Application and Verification, validation threshold temporary relaxation.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
support.investigator — Investigate an unresolved symptom
P Function. Establish the failure and likely domain from current evidence when no existing investigator domain fits; never default an unknown symptom to transfers.
| Contract | Content |
|---|---|
| Input brief | Original request and concrete expected versus observed behaviour; source/actor/environment handles and timestamps Current evidence, failed or stale experience candidates, known uncertainties and the root's bounded read-only question |
| Output brief | Evidence-led failure trace, separating hypotheses and current result or unknown Candidate domain with evidence and a handover brief for the selected specialist; explicit missing evidence/owner action if unresolved |
| Scope and restrictions | One scoped task instance per selected responsibility; a reused profile never shares another case transcript, grants or mutable task state. Boot from the own domain index, task request, pinned profile and confirmed parent plan. Select relevant articles, skills, tools and helpers on demand. Select only the symptom profiles needed by this case. Read-only investigation; no effect applier. Consults may address Configuration or Source but cannot change their systems. |
| Memory / ownership | support/methodologies (read/propose; no new owned memory domain) |
| Role instructions | Anchor the symptom to actual identifiers, times, lifecycle and environment. Trace the observed UI/request failure to the responsible source/data boundary before choosing a domain. Follow current evidence even when the experience search is empty. Compare a same-state valid peer and seek a separating observation; an inaccessible probe is not proof of absence. When a domain fits, return the request, evidence, attempted checks, rejected hypotheses and open questions to the parent for a specialist task. Do not discard those findings or run all specialists blindly. Request manual evidence or an owner-operated action through root where authorised reads cannot settle the issue. No writes, no invented mechanism, no claim of completion from a similarity match. |
| Runtime / connector access | case/paper and permitted source reads scoped connector log/state reads domain-owner consults through parent |
| Context and task continuity | Receives the same relevant case brief as a specialist. Returns a versioned investigation paper so the next investigator continues from evidence instead of restarting from a summary title. |
| May consult | configuration/ipal — What state is this product in? configuration/abacus — Can this be priced as the document says? source/<repository> — What has gone wrong in this service before? A code defect is Hd to Customer-specific development through the root. (a change is Hd) |
| Eval set | support.investigation — declared by the owning module manifest or platform-profiles.json; green before publication (Agents § 0.1) |
| Experience strategy | mechanism; retrieval and index contract |
| Runtime profile identity | support.investigator |
Methods used. common.skill.experience, data.skill.fresh-investigation, common.db-prober.
Source basis. Existing investigation skeleton, Evidence before assumptions, Failed probe is not a finding.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
Domain procedures
data.skill.transfer-diagnosis — Diagnose transfers and families
P Function. Compare source/target lifecycle, mappings, invocation, family members and downstream artifacts to isolate a repairable data mechanism. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Scoped policy/family handles, current state, logs and intended operation |
| Output brief | Mechanism and discriminators: existing/absent target, stuck state, missing/gutted registry, old PDF or uncreated family |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. No blanket retransfer, hardcoded family count or reuse of an old premium ratio. |
| Runtime / connector access | connector.oracle.read connector.logs.read connector.handle-facets; domain ID interpretation domain family assertions (helper candidate) |
Source basis. skill, myr policy missing last row pattern, transfer pipeline, Agents Memory.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.pricing-diagnosis — Diagnose pricing and installments
P Function. Trace requested amount through fed factors, engine output, applied rows and independent installment calculations. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Policy/annex handles, exact pricing invocation, selected covers and current rules |
| Output brief | Wrong-input versus wrong-rating versus persistence/splitter mechanism and component proof |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. POL_PREM_RATE is output, not the input configuration; amounts always carry currency/dimension. |
| Runtime / connector access | connector.logs.read connector.oracle.read connector.abacus.getRates compare current scoped evidence |
Source basis. premium ld applied vs fed factor, pol pplan not a mirror, currency and rates, Agents Memory.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.print-diagnosis — Diagnose document production
P Function. Establish invocation, selected template/condition, data source and generated-document age before changing configuration. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Requested document/operation, exact invocation, template selection and current/persisted data |
| Output brief | Missing invocation, wrong selection, missing data, stale PDF or absent evaluator mechanism |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. IPAL CFG_PRINT_DOCS and INSIS CFG_PRINT_DOC are separate. Producing a new document is an effect. |
| Runtime / connector access | connector.logs.read connector.oracle.read compare current scoped evidence domain family assertions (helper candidate) |
Optional executable helpers. common.tool.aisa-blob. Helper availability and role scope are checked separately.
Source basis. cfg print docs template configuration, insis print configuration, never promote config ahead of evaluator, Agents Memory.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.access-diagnosis — Diagnose account and access layers
P Function. Check master identity, login, actual authentication, office/membership, minimum requested roles and product access independently. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Actor handles, requested action, office/role scope and actual failed login/operation |
| Output brief | Missing-layer and least-scope correction proposal with positive and negative access checks |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. A customer-master row is not a login. Preserve existing roles; no speculative product roles or credential disclosure. |
| Runtime / connector access | connector.oracle.read connector.logs.read connector.handle-facets; domain ID interpretation |
Source basis. skill, ipal user layer separation, identity server ldap authentication, Agents Memory.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.health-diagnosis — Diagnose health lifecycle and totals
P Function. Trace enrolment, claim/report lines, evaluation, amount states and payments on the declared feature environment. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Handle-substituted health request, current schema/version and scoped report/claim evidence |
| Output brief | Mechanism and evidence across item, aggregate and lifecycle state; affected scope and unknowns |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. EU/PII route and special-category minimisation. QA feature evidence is not asserted to exist on PROD. |
| Runtime / connector access | connector.oracle.read connector.logs.read domain health assertions (helper candidate) |
Source basis. health product 8000 anatomy, health claim pipeline step14 overwrites step13, README, Agents Memory.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.add-car — Vehicle nomenclature alignment
P Function. Find the requested make/model in IPAL, INSIS and external nomenclature, reconcile codes/labels and propose only the missing mapping. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Requested make/model/code/blank range, current catalogue and ownership/lifecycle evidence |
| Output brief | Exact missing or inconsistent mapping and scoped admissible change or no-op |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. No global cleanup because one value differs; allocated/used blank ranges require lifecycle proof. |
| Runtime / connector access | Scoped catalogue reads; proposed correction through the effect protocol |
Source basis. Existing add-car procedure.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.add-bso — BSO range and lifecycle check
P Function. Check a named blank range, type, office, overlap and lifecycle; prepare a bounded provisioning packet with counts and post-checks. The same case method defines its evidence, proposed remedy, verification and recovery; investigating, applying and verifying roles consume different authorised segments.
| Contract | Content |
|---|---|
| Input brief | Requested make/model/code/blank range, current catalogue and ownership/lifecycle evidence |
| Output brief | Exact missing or inconsistent mapping and scoped admissible change or no-op |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. No global cleanup because one value differs; allocated/used blank ranges require lifecycle proof. |
| Runtime / connector access | Scoped catalogue reads; proposed correction through the effect protocol |
Source basis. Existing add-bso procedure.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.fresh-investigation — Investigate from fresh evidence
P Function. Run the nine investigation mechanics as phases with a required output each — cross-channel scan, id decoding, invocation before configuration, introspect first, peer baseline, both-systems evidence, verify-don't-guess, intent restatement, environment provenance — and end in one of three states: verified mechanism, containment known, or dead end.
| Contract | Content |
|---|---|
| Input brief | Restated request (object, state, operation), actor/surface/time/environment handles, expected versus observed behaviour Current read scope, negative search coverage and the competing explanations a check must separate |
| Output brief | Evidence-led mechanism with its layer, separating checks, rejected hypotheses and provenance; or explicit unknown with the next evidence needed mechanism_hypothesis {layer, system, domain, confidence} at the hypothesis checkpoint (Stage Investigation § 2) |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. A phase that found nothing is recorded as ran-found-nothing, never skipped; a failed probe is not a finding. Read-only: investigation holds no effect applier and grants no manual database write. |
| Runtime / connector access | connector.oracle.read connector.logs.read connector.handle-facets; domain ID interpretation compare current scoped evidence |
Source basis. investigate-ticket v1 skill, verify, don't guess, DB investigation discipline, a failed probe is not a finding, Stage Investigation.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.framework-sync — Synchronise a cargo framework for child issuance
P Function. Run the IPAL↔INSIS framework checklist for a 1101-type master — policy/annex state, PAS_POLICY_ID mapping, participants, cargo object and AVERAGE_AGENT, covers with the rate triad, loadings, the six date layers and the selector cache — and emit the ordered fix packet for the proven gaps only.
| Contract | Content |
|---|---|
| Input brief | Framework handle, intended child product, current state on both systems and the confirmed master |
| Output brief | Checklist verdict per layer; ordered bounded fix packet with expected counts, or a stop for the owner's decision on a duplicate |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. Never force another transfer; a duplicate proposal or a divergent master is the owner's decision. The end date to extend to must be confirmed: 'synchronise' alone does not say extend, duplicate or repoint. |
| Runtime / connector access | connector.oracle.read compare current scoped evidence domain family assertions (helper candidate) |
Source basis. sync-framework v1 skill, cargo 1101 family sync, framework extension workflow, Operational Method Evidence rows 18 and 20.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.policy-compare — Compare one policy across both administrations
P Function. Walk the living IPAL↔INSIS comparison checklist for one policy — state, annexes, participants, objects, covers, premiums, plans, multi-year family — against a same-lifecycle peer and report the scoped differences; the vehicle part runs the comparator SQL.
| Contract | Content |
|---|---|
| Input brief | Policy handle, environment, the requested invariant and a peer in the same lifecycle state |
| Output brief | Difference report per checklist step with both values; zero-difference steps listed explicitly |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. IPAL POL_PPLAN is an independent splitter, not a mirror of INSIS; a difference there is not automatically a defect. One repaired field is not a repaired policy: the whole requested invariant is re-read. |
| Runtime / connector access | connector.oracle.read compare current scoped evidence tools.run (vehicle comparator) |
Optional executable helpers. data.tool.car-compare-sql. Helper availability and role scope are checked separately.
Source basis. analyze-policy v1 skill, policy sync compare checklist, the checklist is living, vehicle comparator SQL.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.person-policies — List a person's or company's policies
P Function. Answer a scoped lookup — every policy a person or company holds or participates in — from the four-surface SQL template, with provenance and the surfaces that returned nothing; distinguish a role lookup from an arbitrary report.
| Contract | Content |
|---|---|
| Input brief | Authorised person/company handle, requested roles, systems and time coverage |
| Output brief | Scoped table with provenance and limits; duplicates flagged; no raw personal data retained as shared memory |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. Access filtering is enforced by the connector grant, not by the query text. CSV/XLSX rendering is optional mechanical output; the answer still passes the send gate. |
| Runtime / connector access | connector.oracle.read tools.run (SQL template) |
Optional executable helpers. data.tool.person-policies-sql. Helper availability and role scope are checked separately.
Source basis. person-policies v1 skill, person_policies_json.sql, person-policies fast path.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.document-production — Produce or retrieve a policy document
P Function. Resolve the requested document's engine, template, parameters and print gates from the database, retrieve the stored artefact where one exists, and — as a gated effect — request the render through BI Publisher; check the per-document response, never the batch HTTP status.
| Contract | Content |
|---|---|
| Input brief | Policy/annex/bill handle, requested document type and copy variant, current data and template registration |
| Output brief | Stored document bytes with identity, or the gated render request with its parameters and the produced artefact's identity and observed content |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. Producing a new document is an effect with its own gate; retrieving a stored one is read-only. A batch HTTP 200 does not hide a failed document; each document is checked. |
| Runtime / connector access | connector.oracle.read tools.run (aisa-blob) runtime.request-effect (render) |
Optional executable helpers. common.tool.aisa-blob. Helper availability and role scope are checked separately.
Source basis. print-policy v1 skill, policy print recipe, SR_DOCS chain, aisa-blob README.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.health-report-review — Review a health provider report
P Function. Trace a loaded provider report (product 8000) from file to HLT_IMP_DATA lines to claims, evaluations, invoices and payments on the declared environment; classify each line as correct, incorrect or unknown against the check catalogue, minimising special-category data.
| Contract | Content |
|---|---|
| Input brief | Report/file identity, environment, the check catalogue version and the customer's question |
| Output brief | Per-line verdicts with the failing check, linkage errors and the scoped next action; never a claim that upload equals processing |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. QA is the feature environment for product 8000; QA evidence is not asserted to exist on PROD. Special-category data is minimised to handles and counts. |
| Runtime / connector access | connector.oracle.read compare current scoped evidence |
Source basis. review-ao-report v1 skill, report line correction paths, Bulstrad Health KB, product 8000 anatomy.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
data.skill.agent-office — Provision an agent, broker or office
P Function. Walk the INSIS→IPAL provisioning checklist for an exclusive agent, broker or office: the INSIS agent and office records, the IPAL account, branch and role layers, the numbering prerequisites of a new office; report the missing layer and prepare the bounded provisioning packet, never the role grant itself.
| Contract | Content |
|---|---|
| Input brief | Agent or office number, the approved setup and the intended products and office Current INSIS agent/office records, IPAL account, branch, role and numbering state |
| Output brief | Missing-layer verdict with the account tree, and a bounded provisioning packet with counts and post-checks; external Bulstrad IT steps as obligations |
| Scope and restrictions | Selected for the current task when its applicability and published version fit; not loaded or executed merely because the stage exists. The procedure adds no target authority; every effect, including effectful verification, retains its own gate and journal. INSIS agent/office records and IPAL account/branch records are separate provisioning layers; product roles and office numbering are not created by adding the account. A new office's numbering counters are Configuration S5 work when they are configured rather than repaired; the skill reports that boundary. |
| Runtime / connector access | connector.oracle.read connector.handle-facets; domain ID interpretation compare current scoped evidence |
Source basis. add-agent v1 skill, new office provisioning checklist, exclusive agents, office / individual agent activation, Operational Method Evidence row 8.
P Status. planned; seeded from the cited architecture and AISA workspace evidence; no published v2 capability is claimed
Executable helpers used by procedures
data.tool.person-policies-sql — person_policies_json.sql — four-surface policy sweep
P Function. Parametrised SELECT that lists every policy a person or company touches — INSIS policyholder, object participant and creditor roles plus IPAL-only applications — as one JSON document. A query template run by the database connector; deterministic and read-only.
| Contract | Content |
|---|---|
| Input brief | Person or company handle (ЕГН / ЕИК / name / MAN_ID) and the environment |
| Output brief | JSON rows per surface with role, state and identifiers; a surface that returned nothing is reported as such |
| Scope and restrictions | Read-only SELECT; the handle is substituted at the connector so the model never sees the raw identifier (Trust and Data § 4). An empty surface is a negative finding about the probed surfaces, not proof that no relationship exists. |
Executable evidence. SQL template present; v2 package unregistered.
- Source: person_policies_json.sql.
Invocation: SQLcl: @sql/templates/person_policies_json.sql <handle>.
Source basis. person_policies_json.sql, person-policies v1 skill.
P Status. SQL template present; v2 package unregistered.
data.tool.car-compare-sql — srd_integr_compare_ipal_insis_car.sql — vehicle comparator
P Function. Chain-aware comparison of the IPAL vehicle — every OBJ_CAR version through SR_OBJECTS — against the INSIS O_CAR for one policy; returns the differing fields. Vehicles only.
| Contract | Content |
|---|---|
| Input brief | Policy handle and environment |
| Output brief | Field-level differences with both values, or zero relevant differences for the compared fields |
| Scope and restrictions | Vehicle objects only; a full policy-family comparator remains a candidate helper. Read-only; a zero-difference result is evidence for the compared fields, not for the whole policy. |
Executable evidence. Reference SQL present; v2 package unregistered.
- Source: srd_integr_compare_ipal_insis_car.sql.
Invocation: SQLcl: @sql/ref/srd_integr_compare_ipal_insis_car.sql <policy>.
Source basis. COMPARE_IPAL_INSIS_CAR reference SQL, IPAL latest car ↔ INSIS car invariant.
P Status. Reference SQL present; v2 package unregistered.