Operational Ontology Framework epistemic report

Type: agentic-system-analyses/types/agentic-system-epistemic-report.md

Source-and-claim boundary

The question is whether this reference establishes routes from operational information to checked, accepted and later-used content. The scope and exclusions are those of the boundary; evidence identities are in its Source register. SRC-1 supplies doctrine/design for organizational operation and implementation only for publication CI. No operation or causal experiment was supplied. This is a complete public artifact, partial loop, not an inspected production knowledge system.

Direct reads for this member covered both READMEs, all five docs/ files and .github/workflows/public-surface.yml at the frozen checkout, whose HEAD was checked against the reviewed revision. The tracked-file listing agrees with the supplied boundary. Publication policies and configuration outside those reads are covered only by the supplied runtime/boundary, not claimed as direct inspection here. The CLI/dispatch, hook, tool and exposed-operation check found the publication workflow dispatch/push/pull-request surface, not an agent dispatcher; the public-reference exclusion is independently supported by RT-ABS-agent-entry. External kits and private runtime interfaces cannot be inspected within this evidence boundary.

Entry point or operation Source path, SRC-1 Coverage and conclusion limit
Public state and D+L+A model README.md, README.pt-BR.md, docs/dla-model.md Direct; EPI-OBJ-facts, EPI-OBJ-pin, EPI-OBJ-spec, EPI-OBJ-handoff. Roles, not actual state instances.
Weekly client review: Data, Logic, Action, Evidence, Write-back docs/conceptual-example.md Direct; RT-RTE-weekly-review, its separately assessed parts below. No prompts, data or actual reviewer decision.
Source checking and completion docs/governance-principles.md, docs/anti-patterns.md Direct; EPI-RTE-source-check, EPI-RTE-completion-check. No implemented evaluator or test result.
Rule amendment and authority withdrawal docs/governance-principles.md, docs/dla-model.md Direct; RT-RTE-rule-amendment, RT-BAP-action-authority. No mechanical revocation or downstream reload.
Operational usefulness test and context-invention prevention docs/what-is-operational-ontology.md Direct; EPI-CLM-context-invention. Reviewability requirement, no demonstrated benefit.
Publication dispatch and checks .github/workflows/public-surface.yml Direct; RT-RTE-public-surface, EPI-RTE-public-check. Fixed predicates, not operational truth checks.
Contribution/review policy and repository metadata CONTRIBUTING.md, .github/CODEOWNERS, SECURITY.md, .markdownlint.json, NOTICE.md, LICENSE Supplied boundary/runtime only; RT-RTE-public-surface. No additional agent route established there.
Production model calls, hooks, tools, persistence/reload, private evaluations README.md Declared unpublished; RT-ABS-agent-entry. Prevents implementation-wide conclusions on acquisition, acceptance, tool admission or provider learning.
Linked Lite/Starter Kit and external pages README.md Outside frozen source; prevents judging kit prompts/templates and manual outcomes. HTTP-check code does not expose page contents.

The warrant-related claims are EPI-CLM-source-warrant, EPI-CLM-result-evidence, EPI-CLM-context-invention, and the authority claim RT-CLM-human-authority. The public text does not claim an automated research or general truth-production engine. Its stronger goal of preventing invented operational context still needs comparison with the prescribed routes.

Epistemic-object inventory

Fields owned by a record are referenced rather than copied. These are conceptual operative parts, not witnessed artifact instances.

Object ID Candidate truth-apt content or none Claimed role Evidence source ID and local anchor Gap/limit
EPI-OBJ-data see record Data SRC-1, docs/conceptual-example.md Source content and acquisition interface unavailable.
EPI-OBJ-draft see record Draft SRC-1, docs/conceptual-example.md No actual draft; entailment not determinable.
EPI-OBJ-facts see record Facts SRC-1, README.md Metadata requirements are not source verification.
EPI-OBJ-pin see record Pin SRC-1, README.md Identity assertions and prescriptions separated in record; no concrete contents.
EPI-OBJ-spec see record Spec SRC-1, README.md Blocker assertions versus objectives/criteria have different truth aptness.
EPI-OBJ-handoff see record Handoff SRC-1, README.md Reports versus proposed next action separated in record; no actual read-back.

The organizational source after approval is the retained counterpart of Data, not a separately evidenced database. The model's numerical state is not individuated: see RT-CMP-ai. No inference about fixed parameters, online training or knowledge encoded in weights follows from natural-language state roles.

Authority-route ledger

The compact records below hold the ledger fields; shared route records own the substantive findings. An architectural status here is not evidence that a route ran. Weekly-review subroutes preserve the runtime's valid container rather than superseding it.

Route ID: EPI-RTE-data-import Route function: content transformation Architectural status: doctrine only Object/candidate ID: see EPI-RTE-data-import. Content/update relation: see EPI-RTE-data-import. Transition or check target: see EPI-RTE-data-import. Evaluator/condition and domain: see EPI-RTE-data-import. Activation and timing: see EPI-RTE-data-import. Possible or observed result: see EPI-RTE-data-import. Implemented force: see EPI-RTE-data-import. Epistemic authority and scope: see EPI-RTE-data-import. Operational authority: see EPI-RTE-data-import. Behavioral-authority path: see EPI-RTE-data-import. Evidence source ID and local anchor: SRC-1, docs/conceptual-example.md. Claim IDs: EPI-CLM-source-warrant. Mismatch marker: none. Gap/limit: Acquisition and provenance preservation are unspecified.

Route ID: EPI-RTE-draft-generation Route function: content transformation Architectural status: doctrine only Object/candidate ID: see EPI-RTE-draft-generation. Content/update relation: see EPI-RTE-draft-generation. Transition or check target: see EPI-RTE-draft-generation. Evaluator/condition and domain: see EPI-RTE-draft-generation. Activation and timing: see EPI-RTE-draft-generation. Possible or observed result: see EPI-RTE-draft-generation. Implemented force: see EPI-RTE-draft-generation. Epistemic authority and scope: see EPI-RTE-draft-generation. Operational authority: see EPI-RTE-draft-generation. Behavioral-authority path: see EPI-RTE-draft-generation. Evidence source ID and local anchor: SRC-1, docs/conceptual-example.md. Claim IDs: EPI-CLM-context-invention, RT-CLM-human-authority. Mismatch marker: none. Gap/limit: Requested sorting does not establish a semantics-preserving actual draft.

Route ID: EPI-RTE-source-check Route function: check/evidence production Architectural status: doctrine only Object/candidate ID: see EPI-RTE-source-check. Content/update relation: see EPI-RTE-source-check. Transition or check target: see EPI-RTE-source-check. Evaluator/condition and domain: see EPI-RTE-source-check. Activation and timing: see EPI-RTE-source-check. Possible or observed result: see EPI-RTE-source-check. Implemented force: see EPI-RTE-source-check. Epistemic authority and scope: see EPI-RTE-source-check. Operational authority: see EPI-RTE-source-check. Behavioral-authority path: see EPI-RTE-source-check. Evidence source ID and local anchor: SRC-1, docs/conceptual-example.md, docs/governance-principles.md. Claim IDs: EPI-CLM-source-warrant. Mismatch marker: none. Gap/limit: Reviewer competence, evidence consumed and actual accuracy unobserved.

Route ID: EPI-RTE-review-disposition Route function: disposition/acceptance Architectural status: doctrine only Object/candidate ID: see EPI-RTE-review-disposition. Content/update relation: see EPI-RTE-review-disposition. Transition or check target: see EPI-RTE-review-disposition. Evaluator/condition and domain: see EPI-RTE-review-disposition. Activation and timing: see EPI-RTE-review-disposition. Possible or observed result: see EPI-RTE-review-disposition. Implemented force: see EPI-RTE-review-disposition. Epistemic authority and scope: see EPI-RTE-review-disposition. Operational authority: see EPI-RTE-review-disposition. Behavioral-authority path: see EPI-RTE-review-disposition. Evidence source ID and local anchor: SRC-1, docs/conceptual-example.md. Claim IDs: RT-CLM-human-authority. Mismatch marker: none. Gap/limit: No recorded evidence-consuming acceptance decision supplied.

Route ID: EPI-RTE-status-retention Route function: retention Architectural status: doctrine only Object/candidate ID: see EPI-RTE-status-retention. Content/update relation: see EPI-RTE-status-retention. Transition or check target: see EPI-RTE-status-retention. Evaluator/condition and domain: see EPI-RTE-status-retention. Activation and timing: see EPI-RTE-status-retention. Possible or observed result: see EPI-RTE-status-retention. Implemented force: see EPI-RTE-status-retention. Epistemic authority and scope: see EPI-RTE-status-retention. Operational authority: see EPI-RTE-status-retention. Behavioral-authority path: see EPI-RTE-status-retention. Evidence source ID and local anchor: SRC-1, docs/conceptual-example.md. Claim IDs: EPI-CLM-source-warrant. Mismatch marker: none. Gap/limit: Recording is not demonstrated lifecycle integration; no post-acceptance evidence linking or reorganization is specified.

Route ID: EPI-RTE-completion-check Route function: check/evidence production Architectural status: doctrine only Object/candidate ID: see EPI-RTE-completion-check. Content/update relation: see EPI-RTE-completion-check. Transition or check target: see EPI-RTE-completion-check. Evaluator/condition and domain: see EPI-RTE-completion-check. Activation and timing: see EPI-RTE-completion-check. Possible or observed result: see EPI-RTE-completion-check. Implemented force: see EPI-RTE-completion-check. Epistemic authority and scope: see EPI-RTE-completion-check. Operational authority: see EPI-RTE-completion-check. Behavioral-authority path: see EPI-RTE-completion-check. Evidence source ID and local anchor: SRC-1, docs/governance-principles.md. Claim IDs: EPI-CLM-result-evidence. Mismatch marker: none. Gap/limit: No observable result or action-specific measurement procedure supplied.

Route ID: RT-RTE-rule-amendment Route function: behavior/policy adaptation Architectural status: doctrine only Object/candidate ID: EPI-OBJ-pin. Content/update relation: non-truth-apt policy/content update: deliberate amendment of governing rules. Transition or check target: see RT-RTE-rule-amendment. Evaluator/condition and domain: see RT-RTE-rule-amendment. Activation and timing: see RT-RTE-rule-amendment. Possible or observed result: see RT-RTE-rule-amendment. Implemented force: see RT-RTE-rule-amendment. Epistemic authority and scope: No truth license follows from changing an instruction. Operational authority: see RT-RTE-rule-amendment. Behavioral-authority path: see RT-RTE-rule-amendment, RT-BAP-action-authority. Evidence source ID and local anchor: SRC-1, docs/governance-principles.md. Claim IDs: RT-CLM-human-authority. Mismatch marker: none. Gap/limit: No implementation, amendment history or actual withdrawal/rollback supplied. Source precedence is a policy, not cache invalidation.

Route ID: EPI-RTE-public-check Route function: check/evidence production Architectural status: implemented Object/candidate ID: Publication content, see EPI-RTE-public-check. Content/update relation: see EPI-RTE-public-check. Transition or check target: see EPI-RTE-public-check. Evaluator/condition and domain: see EPI-RTE-public-check. Activation and timing: see EPI-RTE-public-check. Possible or observed result: see EPI-RTE-public-check. Implemented force: see EPI-RTE-public-check. Epistemic authority and scope: see EPI-RTE-public-check. Operational authority: see EPI-RTE-public-check. Behavioral-authority path: see EPI-RTE-public-check. Evidence source ID and local anchor: SRC-1, .github/workflows/public-surface.yml. Claim IDs: none. Mismatch marker: none. Gap/limit: Programmatic checks are implemented, but neither actual job results nor mandatory merge gating are supplied.

System-claim versus route comparison

Claim ID Claimed operation or warrant Claim source ID/anchor and evidence layer Doctrine/design support Implemented route IDs Observed-run support Causal support and design limits Supported conclusion Mismatch/unknown
EPI-CLM-source-warrant see record SRC-1, docs/governance-principles.md; doctrine/design see record none for operational facts; EPI-RTE-public-check does not check them None supplied No comparison; metadata and human review do not certify source truth Separates retrieval from truth in doctrine End-to-end source-warrant preservation unknown
EPI-CLM-result-evidence see record SRC-1, docs/governance-principles.md; doctrine/design see record; EPI-RTE-completion-check none for operational completion None supplied No measured action outcomes Explicit evidence requirement, not proof of completion Action-specific evaluators unspecified
EPI-CLM-context-invention see record SRC-1, docs/what-is-operational-ontology.md; doctrine/design see record; EPI-RTE-draft-generation none None supplied No controlled comparison or deployment Concrete missing-status/escalation rules support the intended design Prevention and benefit not established
RT-CLM-human-authority see record SRC-1, docs/governance-principles.md; doctrine/design see record; EPI-RTE-review-disposition none for operational authority None supplied No technical mediation or isolation envelope supplied Human review and draft-only AI prescribed Broad Action vocabulary is not an executable permission set

Bounded conclusion

The reference prescribes acquisition of organizational facts, a prioritized draft, human accuracy review, permission to proceed, and approved-status/blocker recording. It explicitly rejects treating retrieval or fluent model output as evidence of truth or action completion. Acquisition source warrant remains unknown: naming a source and recording confidence do not establish its truth, freshness, or faithful transport. The review policy supplies a plausible place to withhold reliance, not an observed evidence-consuming acceptance decision. Its criterion is accuracy against confirmed organizational information, for weekly-review status and permission to proceed; no acceptance of a produced ampliative theory is established. See EPI-RTE-data-import, EPI-RTE-source-check, EPI-RTE-review-disposition.

The prescribed sorting is a non-ampliative operation if it only reorders unchanged inputs, but neither an actual draft nor a deterministic operational transformer is available. The whole draft's relation is therefore indeterminate, not demonstrated entailed derivation and not established ampliative conjecture. The missing-status rule is a source-supported counterexample to assuming that the design licenses unconstrained inference. The approved-status write-back is a counterexample to assuming that draft-only AI means no organizational write occurs. Retention and next-session continuity are intended; actual read-back, evidence-linked lifecycle integration, and a later consumer's changed behavior remain unestablished. See EPI-RTE-draft-generation, EPI-RTE-status-retention, RT-RTE-weekly-review.

Human review has policy-level withholding and correction force, with the authorized organizational writer as its intended consumer and the next review as its intended horizon. Completion requires observable action evidence, while high-impact effects require human authority. Those are separate epistemic and operational requirements, neither implemented here. Rule amendment can change policy without producing a truth-apt claim; legal/security/revocation precedence is declared but no executable rejection, withdrawal, cleanup or recovery loop is supplied. See EPI-RTE-completion-check, RT-RTE-rule-amendment, RT-BAP-action-authority.

The implemented publication checker produces bounded evidence about exact file/mode/content predicates and HTTP request outcomes. Its nonzero exits have immediate job-continuation force, not established merge-veto force or epistemic authority over operational facts. No target execution was needed for the deterministic code deduction in EPI-RTE-public-check. This positive implementation finding must not be collapsed into the operational loop's doctrine-only status.

No inspected production parameters, traces or comparisons support claims of model adaptation, improved future capacity, autonomous theory building or exercised self-improvement. These remain independent unresolved properties, not a system-wide negative. Human correction of a work product alone establishes none of them. The public artifact's explicitly conceptual scope is not a product failure; it limits which conclusions this run can warrant.

Shared records

Components

none declared in this member

Operative objects

EPI-OBJ-data

Label: Data

Part of: RT-RTE-weekly-review

This object is an input part of the supplied route, not a second route. Candidate truth-apt content: which clients are active, which commitments remain open, deadlines and last confirmed status, scoped to the organization's review. Claimed role: source-of-truth operational input. Representational form and storage: conceptual descriptions in natural language; actual organizational encoding/substrate unspecified. Producer and acquisition path unspecified; AI draft producer and human reviewer are intended consumers. Evidence: SRC-1, docs/conceptual-example.md, Data section. No actual records establish individual propositions or source reliability.

  • active clients;
  • open commitments;
  • agreed deadlines;
  • last confirmed status;
  • named source of truth. --- docs/conceptual-example.md

EPI-OBJ-draft

Label: Prioritized draft

Part of: RT-RTE-weekly-review

This output object is distinct from the route container. Candidate truth-apt content: reported client status and commitments; ordering and recommendations need not themselves be assertions. Claimed role: human-review candidate, not action evidence. Representational form: intended AI-readable/human-readable draft, exact encoding unspecified; storage unspecified. Producer: AI; consumer: reviewer; authority: advisory until review. Evidence: SRC-1, docs/conceptual-example.md; minimum passage also retained in RT-CMP-ai. No populated draft permits proposition-level entailment analysis.

EPI-OBJ-facts

Label: Facts

No counterpart object is declared in the supplied runtime; unlike its weekly-review route this is the general long-term artifact role. Candidate truth-apt content: long-term observations over their source/date scope. Representational form: conceptual natural-language role, not a schema; actual encoding/substrate unknown. Claimed role: source/date/confidence-bearing retained observations; producer, check procedure and actual later consumer unspecified. Evidence: SRC-1, README.md.

| Facts | Long-term observations with source, date and confidence | Medium | --- README.md

EPI-OBJ-pin

Label: Pin

Closest supplied record: RT-RTE-rule-amendment, which describes change, not this artifact. Candidate truth-apt part: identity/domain assertions if instantiated. Non-truth-apt part: non-negotiable prescriptions. These have different semantic roles; neither concrete content nor an implemented consumer is supplied. Representational form: conceptual natural-language state role; storage unspecified. Intended operational authority belongs to rules, not automatically to identity assertions. Evidence: SRC-1, README.md; precedence limits in docs/governance-principles.md are retained by the supplied route.

| Pin | Stable identity, domain boundaries and non-negotiable rules | Low | --- README.md

EPI-OBJ-spec

Label: Spec

No counterpart object is declared in the supplied runtime. Candidate truth-apt part: blocker/task-state assertions if instantiated. Objectives, tasks as prescriptions and acceptance criteria specify desired outcomes rather than their achievement. Representational form: conceptual natural-language role; actual storage unspecified. Claimed role: current work guidance; producer, verifier and concrete consumer unavailable. No claim that a criterion was satisfied follows from storing it. Evidence: SRC-1, README.md.

| Spec | Current objective, tasks, blockers and acceptance criteria | High | --- README.md

EPI-OBJ-handoff

Label: Handoff

Part of: RT-RTE-weekly-review

This is the intended continuity output object of that route; the README supplies its more general role. Truth-apt part: reports of decisions, attempts and results. Non-truth-apt part: proposed next action. Representational form: conceptual natural-language continuity record; storage unspecified. Claimed role: append-only session continuity. Immediate reviewer/next operator are intended consumers; no actual artifact or read-back demonstrates lineage, acceptance or activation. Evidence: SRC-1, README.md, docs/conceptual-example.md.

| Handoff | Session continuity: decisions, attempts, results and next action | Append-only | --- README.md

Routes

EPI-RTE-data-import

Label: Review data import

Part of: RT-RTE-weekly-review

Endpoints/progression: organizational source → EPI-OBJ-data available for the AI/reviewer. Owner: organization; physical acquisition actor unspecified. Function: acquisition of review inputs before drafting. Content/update relation: truth-apt transformation: acquisition/import; source warrant unknown. Check target: none at this transport stage; it does not itself evaluate source reliability. Result: candidate inputs for review, with no independent truth license. Behavioral-authority path: AI receives operational information by an unspecified channel, advisory context for the current draft. No implemented force or demonstrated source-preservation guarantee. SRC-1, docs/conceptual-example.md, Data passage in EPI-OBJ-data.

  • Immediate return: Conceptually available client/commitment/deadline/status data.
  • Later read-back: uninspected — no acquisition or reload implementation supplied.
  • Delegated visibility: AI and reviewer are intended recipients; no worker delegation described.
  • Selection predicate: Active clients and open commitments; physical query unspecified.
  • Invalidation or expiry: uninspected — no input expiry procedure published.
  • Activation or effect: Intended drafting input; no actual delivery/behavior trace.
  • Evidence limits: Data list, not connector or observation.
  • design conclusion status: claimed
  • implementation conclusion status: uninspected
  • operation conclusion status: uninspected

EPI-RTE-draft-generation

Label: Prioritized draft generation

Part of: RT-RTE-weekly-review

Endpoints/progression: EPI-OBJ-data and Logic → AI-produced EPI-OBJ-draft → reviewer. Owner: AI draft role under human-supervised routine. Content/update relation: truth-apt transformation: indeterminate. The requested ordering alone would be non-ampliative reshaping if inputs are preserved, but the text does not determine the propositions in an actual draft. Entailed derivation versus unsupported inference cannot be decided without inputs, prompt/transformer and output. Lack of entailment evidence does not establish non-entailment. Target/condition: overdue first, missing status flagged rather than inferred, completion only with recorded confirmation, disputed/sensitive items escalated. Domain: weekly client review. These are instructions, not observed evaluator results. Possible result: prioritized review candidate; no independent epistemic license. Operational path: AI → draft channel → reviewer, advisory for current review; no AI client contact or source update in this example. Implemented force: none found for these instructions. SRC-1, docs/conceptual-example.md.

  • Immediate return: Prioritized draft for human review.
  • Later read-back: inapplicable — this subroute produces a current-review candidate; continuity is a separate retention function.
  • Delegated visibility: Human reviewer receives the draft by prescription; interface unspecified.
  • Selection predicate: Overdue before future commitments; missing, sensitive and disputed statuses flagged/escalated.
  • Invalidation or expiry: uninspected — draft versioning/replacement not defined.
  • Activation or effect: Drafting prescribed, not observed.
  • Evidence limits: Non-executable example; no concrete input/output pair or model configuration.
  • design conclusion status: claimed
  • implementation conclusion status: uninspected
  • operation conclusion status: uninspected
  • overdue commitments appear before future commitments;
  • missing status is labeled for verification rather than inferred;
  • no item is marked complete without recorded confirmation;
  • sensitive or disputed items are escalated to a human. --- docs/conceptual-example.md

EPI-RTE-source-check

Label: Review accuracy check

Part of: RT-RTE-weekly-review

Endpoints/progression: draft candidate → human accuracy review → judgment of accuracy/correction need. Owner/evaluator: human reviewer. Target/domain: client/commitment/status propositions for this weekly review, against confirmed organizational information. Content/update relation: no content change in the checking function; correction is a possible subsequent disposition. Possible results: accurate, requires correction, disputed/escalated. Behavioral-authority path: reviewer judgment → review decision → intended correction/withholding, policy force within this review. No implemented force; truth license is bounded to items actually confirmed against adequate evidence, not reviewer infallibility. Actual consumed evidence is unobserved. Proposal: AI draft; human decides/vetoes; answer reference is named source/recorded confirmation, not an expected-answer dataset or model oracle. Operating mode: recurring human-supervised routine; trigger: draft and uncertain items, not capacity optimization. SRC-1, docs/conceptual-example.md, docs/governance-principles.md.

  • Immediate return: Intended item-level accuracy/correction judgments.
  • Later read-back: inapplicable — checking alone does not persist the judgment.
  • Delegated visibility: Reviewer sees draft; no computational reviewer or sub-agent supplied.
  • Selection predicate: Draft items, especially missing/sensitive/disputed status.
  • Invalidation or expiry: uninspected — recheck cadence and evidence-age thresholds unspecified.
  • Activation or effect: Intended input to approval/correction; no recorded review supplied.
  • Evidence limits: Reviewer responsibility is specified, actual evidence consumption is not demonstrated.
  • design conclusion status: claimed
  • implementation conclusion status: uninspected
  • operation conclusion status: uninspected

The reviewer confirms which items are accurate, which require correction and which may proceed. --- docs/conceptual-example.md

EPI-RTE-review-disposition

Label: Review disposition

Part of: RT-RTE-weekly-review

Endpoints/progression: accuracy judgment → human approval, correction or withholding → approved status for authorized recording. Owner: reviewer; actual writer unspecified. Object: EPI-OBJ-draft. Content/update relation: no content change for approval/withholding itself; corrected propositions are not supplied, so their semantics cannot be assessed. Target: whether proposed review status is accurate and which items may proceed; domain/intended use: weekly review and authorized status recording. Evaluator: human, guided by recorded confirmation and source-of-truth information. Admission criterion is named in doctrine but no evidence-consuming decision record exists here. No produced accepted ampliative claim is established. Operational path: reviewer → approval/correction communication → authorized writer, permissive or withholding by policy for this review; implemented force not supplied. Distinguish epistemic approval of accuracy from operational permission to proceed; one is not proof of the other. No software veto, cleanup procedure or guaranteed rejection channel is supplied. SRC-1, docs/conceptual-example.md; passage in EPI-RTE-source-check and write-back passage in EPI-RTE-status-retention.

  • Immediate return: Intended approval, correction requirement or permission decision.
  • Later read-back: inapplicable — retention is separately assessed.
  • Delegated visibility: Authorized organization-side writer is the intended consumer; no identity/interface supplied.
  • Selection predicate: Reviewer-confirmed items, with no completion absent recorded confirmation.
  • Invalidation or expiry: uninspected — reversal/rejection log and approval expiration unspecified.
  • Activation or effect: Intended withholding/correction or permission to record approved status; no actual decision supplied.
  • Evidence limits: Human responsibility is doctrine, not a recorded acceptance instance or technical gate.
  • design conclusion status: claimed
  • implementation conclusion status: uninspected
  • operation conclusion status: uninspected

Proposal source: AI draft; decision/veto: human reviewer. Operating mode: manual recurring review. Answer reference: confirmed organizational records, not an automated oracle. Trigger/change: review of proposed statuses; guidance: the four Logic rules. What persists is assigned to the following retention route. Correction/escalation are prescribed; rollback of an already-written update is unspecified.

EPI-RTE-status-retention

Label: Approved status retention

Part of: RT-RTE-weekly-review

Endpoints/progression: approved status and remaining blockers → authorized organizational source; next review point → EPI-OBJ-handoff. Owner: authorized organization-side process, physical writer unspecified. Function: retention, not established lifecycle integration. Object: status counterpart of EPI-OBJ-data plus handoff. Content/update relation: truth-apt transformation: indeterminate for the combined recording/summarizing step; exact status transfer could preserve content, but handoff selection and preservation are unspecified. No evaluator of fidelity is defined. Possible result: recorded status, blockers and next-review pointer. Epistemic authority: recording grants no new truth warrant; intended reliance is on the preceding human review. Operational path: organizational source/handoff → next operator via unspecified read-back channel, intended guidance across sessions. No implemented force, lineage guarantee or actual later use. SRC-1, docs/conceptual-example.md, docs/governance-principles.md.

  • Immediate return: Intended approved-status/blocker record and short handoff.
  • Later read-back: uninspected — next-review continuity is intended but delivery and consumption are not supplied.
  • Delegated visibility: Organization and next operator are intended recipients; access controls unspecified.
  • Selection predicate: Approved status, remaining blockers and next review point.
  • Invalidation or expiry: uninspected — status replacement/expiry and handoff supersession not defined.
  • Activation or effect: Durable recording prescribed; no write or later-use trace supplied.
  • Evidence limits: No storage schema, actual artifact, lineage links or later consumer invocation.
  • design conclusion status: claimed
  • implementation conclusion status: uninspected
  • operation conclusion status: uninspected

Admission: human-approved statuses; writer not specified. Trigger/change: post-review recording; guidance: review rules and Write-back. Rejection before recording belongs to reviewer; rollback after recording unspecified. No additional fidelity oracle or automatic fact admission exists in the published example.

The approved status and remaining blockers are recorded in the organization's authorized source. A short handoff preserves the next review point. --- docs/conceptual-example.md

EPI-RTE-completion-check

Label: Completion evidence check

No counterpart route is separately declared in the supplied runtime; its weekly route quotes a completion-specific instance, but this record covers the general action-result requirement. Endpoints/progression: action completion claim → appropriate observable result → intended completion determination. Owner/evaluator: responsible operator unspecified at the general level. Target/domain: whether a named action occurred and produced its appropriate result, not whether the model sounded convincing. Object: completion assertion; actual instance unavailable. Content/update relation: no content change in the checking requirement. Possible result: sufficient or insufficient evidence; no judgment observed. Epistemic license: only the action/result actually evidenced, not generating process, causal mechanism, transfer or safety. Operational path: result evidence → operator's completion decision, normative withholding of completion until verification, current action horizon. Implemented force: none supplied. SRC-1, docs/governance-principles.md, docs/anti-patterns.md.

  • Immediate return: Intended evidence-based completion assessment; no result supplied.
  • Later read-back: inapplicable — the checking rule alone does not persist outcomes.
  • Delegated visibility: uninspected — concrete evidence channels and recipients unspecified.
  • Selection predicate: Observable result appropriate to the action.
  • Invalidation or expiry: uninspected — result validity horizon not specified.
  • Activation or effect: Completion is normatively withheld absent evidence, with no implemented guard or observed decision.
  • Evidence limits: No action-specific measurement, expected outcome, trace or evaluation artifact.
  • design conclusion status: claimed
  • implementation conclusion status: uninspected
  • operation conclusion status: uninspected

Producing an answer is not evidence that work happened. Completion requires an observable result appropriate to the action. --- docs/governance-principles.md

EPI-RTE-public-check

Label: Publication predicate checks

Part of: RT-RTE-public-surface

This is the evaluator part of the supplied publication route, not a second CI system. Endpoints/progression: checked-out tracked files → exact allowlist/mode/marker/regex predicates; fixed URL requests → job-step outcome. Owner: repository workflow, executed by GitHub Actions if triggered. Target/domain: equality to a 14-path list, tracked modes, detectable credential strings, implementation-like Markdown lines, and HTTP request success. Content/update relation: no content change to organizational knowledge; diagnostic assertions are entailed derivations within the inspected predicates, conditional on actual inputs and execution. Evaluator: Bash/Python/curl, not a semantic truth evaluator. Timing: dispatch, main push or main-targeted pull request. Possible result: nonzero failure or continuation; no observed result. Implemented force: exits fail steps and constrain job continuation. Behavioral-authority path: step exit → GitHub Actions job status, enforcing within this workflow invocation; maintainer/merge consumption is not established. License: only these syntactic/request predicates; not absence of all credentials/code, page correctness, public-document truth or operational safety. SRC-1, .github/workflows/public-surface.yml.

Concrete deduction, not execution: if a tracked UTF-8 README.md contained a line def helper():, the anchored def branch would match; .md suffix passes the condition; the script would append implementation-like content: README.md; the nonempty error list would trigger SystemExit(1). The corresponding line in a non-Markdown file does not take that regex branch. This countercase bounds the check's coverage. A successful HTTP response also does not check the page's claims. Failure is diagnostic/rejection of the job's predicate, not deletion, cleanup or rollback of repository content.

  • Immediate return: Conditional CI diagnostics and success/failure status.
  • Later read-back: inapplicable — no agent-retained content or later agent consumer is implemented here.
  • Delegated visibility: Checkout/job reads repository; fixed remote sites receive requests, not operational state.
  • Selection predicate: Tracked files, exact paths/modes, credential marker substrings, Markdown regex and three fixed URLs.
  • Invalidation or expiry: Each trigger checks its checkout; no retained operational candidate.
  • Activation or effect: Nonzero exits wired to step failure; execution and mandatory merge blocking unobserved.
  • Evidence limits: No supplied CI run, branch protection, page content or maintainer admission decision.
  • implementation conclusion status: wired
  • operation conclusion status: uninspected

For publication admission, see RT-RTE-public-surface: humans propose/admit public content; this subroute is an automated check, not the merge decider. Answer references are hardcoded predicates and HTTP outcomes. No capacity-improvement objective, candidate optimizer or operational fact acceptance is present in this implementation.

          if path.suffix == ".md" and implementation.search(text):
              errors.append(f"implementation-like content: {path_text}")

--- .github/workflows/public-surface.yml

      if errors:
          print("\n".join(errors), file=sys.stderr)
          raise SystemExit(1)

--- .github/workflows/public-surface.yml

Claims

EPI-CLM-source-warrant

Label: Source before retrieval

Distinct from RT-CLM-human-authority: this is the fact-reliance rule, not action authorization. Claimed operation/warrant: retrieval alone does not determine truth; relevant facts require source, freshness expectation and owner. Source: SRC-1, docs/governance-principles.md; reinforced by docs/dla-model.md. Doctrine/design only. It prescribes warrant discipline without claiming metadata suffices to establish truth.

Finding information is not the same as determining truth. Every relevant fact needs a source, freshness expectation and owner. --- docs/governance-principles.md

EPI-CLM-result-evidence

Label: Verification before completion

Distinct from RT-CLM-human-authority: this is a completion-evidence condition, not permission. Claimed operation/warrant: completion requires observable action-appropriate results, not an answer. Source: SRC-1, docs/governance-principles.md; minimum passage in EPI-RTE-completion-check. Doctrine/design only; not a claim that any action actually completed.

EPI-CLM-context-invention

Label: Prevent invented context

Distinct from RT-CLM-human-authority: this concerns invented operational context, not the grant set. Claimed goal: prevent fabricated context in consequential decisions. Minimum operational-usefulness test requires a reviewer to identify source, rule, authority, result evidence and state for the next session. Source: SRC-1, docs/what-is-operational-ontology.md; doctrine/design, not observed prevention or causal benefit.

The goal is not to make an AI system sound informed. The goal is to prevent it from inventing operational context when decisions have consequences. --- docs/what-is-operational-ontology.md

A model is operationally useful only when a reviewer can determine:

  • which source informed the decision;
  • which rule was applied;
  • which authority permitted the action;
  • which evidence shows the result;
  • which state must be preserved for the next session. --- docs/what-is-operational-ontology.md

Evidenced absences

none declared in this member

Behavioral-authority paths

none declared in this member