docs(agents): canonicalize OTCLIENT-TIBIA-RE in OTClient

[blakinio]

Objective

Make blakinio/otclient the single canonical repository/coordination surface for OTCLIENT-TIBIA-RE, route the alias through the dedicated OTClient runner model, and import/index the material historical Oteryn evidence needed for continuation.

Scope

Exactly seven documentation paths:

• docs/agents/SHORT_COMMANDS.md — explicit OTCLIENT-TIBIA-RE registry;
• docs/agents/prompts/OTCLIENT_TIBIA_RE_CANONICAL.md — additive wrapper over the existing programme prompt;
• canonical state, imported login-recovery evidence and external-source migration manifest;
• prompt-eval/manual scenario matrix report;
• bounded consolidation task record.

Runtime ownership

New live work belongs in blakinio/otclient and uses [self-hosted, otclient, synology] during migration, then the stricter tibia-re label after PR #280 is redeployed. Historical oteryn-staging, oteryn-synology-staging, oteryn-tibia-client-analysis and /var/lib/oteryn-staging-state/** are evidence only, not active dependencies.

All material findings, experiments, failures, artifact references and next_action state must be persisted or indexed in blakinio/otclient before return/rotation.

Related implementation lanes

• ci: prove official Tibia login without OCR on Synology #48 runtime/login/live experiments
• feat(tools): add fail-closed worldmap reconstruction pipeline #279 worldmap/OTBM pipeline
• ci: add dedicated OTClient and OTS runner stack #280 dedicated runner infrastructure
• feat(tools): add exact-version Tibia runtime bridge #283 stable runtime bridge

This PR does not edit their owned implementation paths.

Prompt evaluation claim boundary

The repository has no executable fresh-agent behavioral eval harness for this wrapper, and no owner-funded Codex/API/model quota was used. OTCLIENT-20260813-tibia-re-canonical-prompt-eval.md therefore records:

• a 14-case manual scenario matrix whose candidate rules are statically represented consistently;
• deterministic verification of the resulting Git/PR/runtime-routing state;
• behavioral_agent_trials: NOT_RUN and behavioral_candidate_status: NOT_EVALUATED.

It explicitly does not claim an automated/model-behavior 14/14 PASS.

Review findings addressed

• P1 prompt-eval overclaim: corrected to the manual-matrix/behavioral-not-evaluated boundary above.
• P1 missing task coordination fields: modules_touched, depends_on, blocks, and cross_repository_tasks are explicitly declared.

Boundaries

No external repository writes, no Codex/OpenAI API/owner-funded AI quota, no credentials, no proprietary client bytes.

Validation

• changed-file inventory: exactly seven task-scoped documentation files;
• earlier exact-head CI run 31680504103 passed on head 289fa706b3edd5547ac9d561731cb9400776053d, including CI / Required job 94385013470;
• current head after review repairs: b2089d6f1fb9dcdd78acd67f2c9cb132fa42738f;
• final CI on that exact head is required before merge;
• runtime deployment remains owned by ci: add dedicated OTClient and OTS runner stack #280/ci: prove official Tibia login without OCR on Synology #48 and is explicitly not claimed complete by this documentation PR.

Task

OTC-20260813-tibia-re-canonical-runtime

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 95f562d0b4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".