mj-agentlab/html-guide-kit
v0.1.36MIT
Create a new interactive HTML teaching guide from supplied material after the user confirms its reader, task and understanding questions. Deliver guide text in chat.
Changelog
[Unreleased]
[0.1.36] - 2026-10-15
- HTML Guide Kit 0.1.36 changes one existing framework paragraph: distinguish current one-step permission or result from actual claims about multistep paths or unreachability. Only for the latter, check complete source actions, resets and compatible successors within the actual scope, applying each step premise.
- Missing premises block only dependent conclusions and do not erase other known paths. Do not assume changed conditions or successful outcomes; unshown teaching is not absent source information. Narrow unsupported blanket negatives without requiring exhaustive path enumeration. Entry and interaction-design remain exact Source35 bytes; the public contract and other methods stay intact.
- Preserve Source35 original whole failure, unrun inputs and limited native client evidence. Fresh engineering, canonical installation and independent behavior/client acceptance remain required. Refs #217, #220.
[0.1.35] - 2026-10-15
- HTML Guide Kit 0.1.35 changes only two existing framework spans: identify the actual object set and judgment dimension, cross-check related source units per object, and preserve known permission, prohibition and results before stating remaining unknowns.
- Unshown or unchecked cases are not unprovided source rules. If the object set was not fully checked, narrow the statement to checked items and evidence-bound unknowns. Entry and interaction-design remain exact Source34 bytes; all prior methods and the public contract stay intact.
- Preserve Source34 original whole failure, unrun inputs and limited native pending-stage evidence. Fresh engineering, canonical installation and independent behavior/client acceptance remain required. Refs #217, #220.
[0.1.34] - 2026-10-15
- HTML Guide Kit 0.1.34 establishes necessary shared example inputs, actual coverage and local overrides before giving results; independent examples receive their required starting inputs.
- Audit complete local source/design action attribution, including source non-field effects; do not restrict source restoration to terminal values. Keep explicit full-unit reuse and dimension-bound properties.
- Apply only three reviewed entry/framework spans and keep interaction-design exact. Preserve Source33 whole failures and unrun inputs, all original criteria, broad permissions, conditional dependencies and public contract. Fresh engineering, installation and independent behavior/client evidence remain required. Refs #217, #220.
[0.1.33] - 2026-10-15
- HTML Guide Kit 0.1.33 preserves all parallel source actions, objects and non-field effects before terminal field values; field-only summaries cannot replace a complete operation or recovery unit.
- Apply only three reviewed entry/framework spans; keep interaction-design exact, actual local scope reuse, goal-based omitted units and dimension-bound necessary properties.
- Preserve original0.1.32 whole failure, unrun cases and earlier limited client evidence. Fresh engineering, canonical installation and independent behavior/client evidence remain required. Refs #217, #220.
[0.1.32] - 2026-10-15
- HTML Guide Kit 0.1.32 preserves condition conclusions decided by known AND/OR operands even when another input is unknown, using the complete source criterion and actual dependencies.
- Apply only three reviewed entry/framework spans; retain interaction-design bytes, source scope, necessary-versus-sufficient distinctions, actual-example inputs and explicit Read-only new-guide/text boundaries.
- Preserve original0.1.31 whole failure, partial/unrun cases and prior limited client results. Require fresh engineering, canonical installation and independent behavior/client evidence. Refs #217, #220.
[0.1.31] - 2026-10-15
- HTML Guide Kit 0.1.31 separates source prerequisites/initial values from actual example inputs, using only explicitly applicable shared setups or clearly marked teaching assumptions. Candidate questions must preserve source operation permission and result distinctions.
- Apply only three reviewed entry/framework spans; retain interaction-design bytes, known broad permissions, relevant unknown dependencies and the explicit Read-only new-guide/text contract.
- Preserve actual0.1.30 whole failures, partial/unrun paths and limited client results; require fresh engineering, precise installation and independent behavior/client checks. Refs #217, #220.
[0.1.30] - 2026-10-15
- HTML Guide Kit 0.1.30 checks the actual source scope and all stated prerequisites before declaring information gaps. Preserve covered operation permissions and explicit results; unmentioned repetition, previous field values or extra history do not create new prerequisites.
- Apply only the approved entry/framework spans. Keep interaction-design byte-exact, genuine unknown outcomes and scope boundaries, the explicit Read-only text contract, target confirmation and teaching methods.
- Preserve the real0.1.29 cancellation-permission failure and all partial/unrun evidence. Fresh engineering, canonical installation, independent behavior and current client checks remain required. Refs #217, #220.
[0.1.29] - 2026-10-15
- HTML Guide Kit 0.1.29 puts single-file reading and evidence-based refusal reporting before the first read, and aligns the framework: unspecified causes and permission scope stay unknown; later successful reads do not prove earlier refusal causes.
- Apply only the approved entry/framework spans; interaction-design remains byte-exact. Preserve the explicit Read-only chat-text contract and all target confirmation, source fidelity, simulation, recovery and teaching feedback boundaries.
- Preserve original0.1.28 failures. Fresh canonical installation, independent behavior and this version’s client checks remain required; explicit high stays separate from the actual low default. Refs #217, #220.
[0.1.28] - 2026-10-15
- HTML Guide Kit 0.1.28 condenses the entry and two references into shared ordered evidence, target confirmation, complete source rules, example computation and whole-guide checking.
- Preserve the explicit Read-only chat-text contract, adaptive depth, conditions, restoration, bounded unknowns and teaching feedback/retry without scores or production steps; no fixture answers or fixed guide structure.
- Original0.1.27 failures remain separate. Require fresh canonical installation, behavior and client checks; external explicit reasoning selection stays separate from the real default. Refs #217, #220.
[0.1.27] - 2026-10-15
- HTML Guide Kit 0.1.27 classifies each source action or field result and added instructional result before composing sentences or lists; a grouped source citation cannot cover added design items.
- Bound additional-path gaps separately from known direct-operation judgments. Only three existing spans in SKILL.md and interaction-design.md change; guide-framework.md and the public contract remain exact.
- Preserve source0.1.26 actual failures and partial/not-run paths; require fresh installation and frozen behavior/client acceptance. Refs #217, #220.
[0.1.26] - 2026-10-14
- HTML Guide Kit 0.1.26 matches the actual object, state, direct operation and result dimension against permissions, prohibitions, restoration and exceptions before identifying a specific missing dimension, including proposal summaries.
- Only two existing spans in SKILL.md and guide-framework.md change; interaction-design.md, contract, icon, policy and other packages remain. Preserve source0.1.25 actual fidelity failure, legitimate comparison notes and unexercised paths; require fresh installation and frozen behavior/client checks. Refs #217, #220.
[0.1.25] - 2026-10-14
- HTML Guide Kit 0.1.25 reorganizes SKILL.md and both references into an ordered method. Check known rules before gaps, then compute each example’s actual missing conditions and compare the feedback with its inputs.
- Preserve the explicit Read-only new-guide contract, source fidelity, confirmation, simulation, self-check and text boundaries. No fixed guide structure or fixture answers. Preserve actual0.1.24 failures; fresh installed behavior and client checks remain required. Refs #217, #220.
[0.1.24] - 2026-10-14
- HTML Guide Kit 0.1.24 checks exclusive claims against the same result dimension and separates operation permission from achieving a target result. Incomplete local relationships can use faithful text or complete rule units without a fixed format.
- Only two references change in three existing spans; entry, icon, public contract and other plugins remain. Preserve source0.1.23 actual failures, passes and unexercised paths; require fresh exact installation and behavior/client verification. Refs #217, #220.
[0.1.23] - 2026-10-14
- HTML Guide Kit 0.1.23 reorganizes source-rule checks and two-way guide audits, retaining all source recovery actions in local relations and keeping teaching input choices separate from source limits. The entry and public contract remain unchanged.
- Preserve source0.1.22 original failures and actual merged-version installation/client evidence; require fresh installation and frozen behavior verification. Refs #217, #220.
[0.1.22] - 2026-10-14
- HTML Guide Kit 0.1.22 preserves every source prerequisite in result-bearing graph connections, including required current state and direct operation, or explicitly establishes the actual diagram scope. Pure condition relations remain distinct from operation results.
- Preserve source0.1.21 13 PASS / 1 FAIL in 30 complete core turns; static review is not fresh installation, model or client acceptance. Refs #217, #220.
[0.1.21] - 2026-10-14
- HTML Guide Kit 0.1.21 retains the current state and all prerequisites in permission labels and short feedback; writing requirements do not establish a local scope. Pure condition judgments retain their own boundary.
- Preserve the full public contract and actual source0.1.20 failure, invalid captures and limited native paths; require fresh installation and frozen behavior acceptance. Refs #217, #220.
[0.1.20] - 2026-10-14
- HTML Guide Kit 0.1.20 matches the complete source-rule range before marking a permission or field unknown, and budgets headings, labels, prompts and text against explicit user limits with room for review.
- Preserve the complete public contract, source0.1.19 core passes, three actual extended failures and unexecuted paths. Fresh installation and frozen behavior verification remain required. Refs #217, #220.
[0.1.19] - 2026-10-14
- HTML Guide Kit 0.1.19 arranges each operation and result dimension separately, preserving known prohibitions beside unspecified fields. Bind source recovery and each added teaching action/result to their own identity, basis or design reason and scope; check the complete guide before delivery.
- Preserve the complete contract and valid shared-rule abbreviations through four existing paragraph changes, retaining original source0.1.18 failures and requiring fresh installation and frozen behavior checks. Refs #217, #220.
[0.1.18] - 2026-10-14
- HTML Guide Kit 0.1.18 identifies the source of each trigger, operation, action, object, field and result. A new teaching trigger is explicitly marked as design beside the trigger; reused source recovery actions and results retain their own source citation. Equal results do not establish equal triggers.
- Preserve all existing contracts and methods with one small change per runtime resource, retaining source0.1.17 failures and requiring fresh installation and frozen behavior checks. Refs #217, #220.
[0.1.17] - 2026-10-14
- HTML Guide Kit 0.1.17 reads one source or fragment at a time without bundling unrelated file inspection or encoded scripts. Respect explicit access denials; generic tool rejection does not establish a permission cause.
- Explain unsupported tasks using the known Read/chat/new-guide contract. Preserve necessary source conditions throughout summaries and author checklists, and bind each source or teaching claim explicitly.
- Keep both references byte-identical to0.1.16; preserve the original failed behavior and require fresh installation and frozen behavior checks. Refs #217, #220.
[0.1.16] - 2026-10-14
- HTML Guide Kit 0.1.16 reports known read-only mode separately from the cause of a particular denied read. A generic policy response does not identify the policy type or permission range.
- An explicit selection of an existing target is sufficient; a modified target needs confirmation of its adoption.
- Consolidate duplicated reference rules; preserve the source, confirmation, simulation, teaching feedback, recovery and text-only boundaries.
- Retain source0.1.15 nine real sessions/sixteen inputs5PASS/3FAIL/1INVALID_RUN, actual missing reference lines and the unknown capture cause. Fresh installation and behavior remain required. Refs #217, #220.
[0.1.15] - 2026-10-14
- HTML Guide Kit 0.1.15 reports only evidenced denied-read facts and keeps unverified causes or permission ranges unknown; scope refusal does not need an extra environment diagnosis.
- Complete source results include only explicitly supported field results. Do not infer unchanged values for unmentioned fields; separate any local teaching input persistence as design.
- Retain0.1.14 actual10 sessions/21 inputs6PASS/3FAIL/1UNEXERCISED, original read facts and pause races. Fresh canonical installation and behavior remain required; native clients and reality effects remain separate. Refs #217, #220.
[0.1.14] - 2026-10-13
- HTML Guide Kit 0.1.14 presents complete source rules and an explicit shared applicability convention before rewritten operation results, including summaries and author checklists. Completion words do not supply missing source premises.
- Separate UI design rationale from source normative support; state actual received read coverage and verified failures without inventing path-policy causes. Explicit early reference-reading requests preserve the target-confirmation gate.
- Retain0.1.13 actual12 sessions/24 inputs and6PASS/4FAIL/2UNEXERCISED; native/window/effect limits remain separate. Fresh canonical installation and behavior remain required. Refs #217, #220.
[0.1.13] - 2026-10-13
- HTML Guide Kit 0.1.13 composes source rules with all explicit actions, objects, field results and exceptions; full recovery and exclusive relations preserve recovery rules. Define each attempt’s entry, current state, retry, continuation and reset before feedback.
- Preserve local scope and source/UI attribution, explicit confirmation, Read capability and text-only new-guide boundary. Static review does not substitute fresh installation or behavioral acceptance.
- Retain 0.1.12 independent failures and exact original evidence. Refs #217, #220.
[0.1.12] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.12 derives each operation statement from a complete source rule, retaining local premises in examples, diagrams and summaries rather than freely shortening results.
- Judge one direct operation in its current state at a time; unknown later paths do not override source-covered allowed or prohibited direct results.
- Preserve every actual0.1.11 call, localPASS and finalFAIL. Require fresh installed acceptance with unchanged inputs, criteria and model defaults. Refs #217, #220.
[0.1.11] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.11 preserves complete operation conditions in every overview, visible conclusion, retry prompt and detailed result, using explicitly scoped common rules where appropriate.
- Attribute source reset fields and additional display or feedback clearing separately at each item; a shared source citation cannot authorize new interface behavior.
- Preserve all actual0.1.10 reviews, including the initialB6.1PASS and the same-attempt supplementalFAIL. Require fresh installed acceptance with unchanged inputs, criteria and model defaults. Refs #217, #220.
[0.1.10] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.10 declares each feedback block's current operation/state scope across all its branches; re-evaluate after state changes, or use explicitly reset independent scenarios without inventing source rules.
- Preserve existing reader/task clues when a complete target or criterion is missing. Scope refusals must not add unsupported claims about absent source fields.
- Retain every actual0.1.9 failure and incomplete/unrun path; require fresh installed acceptance with unchanged inputs and criteria. Refs #217, #220.
[0.1.9] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.9 separates execution references from user-designated teaching material, checks broad source rules before declaring gaps, and uses current premises after each state-changing interaction.
- Preserve source-covered local inputs despite an unknown complete domain. Require evidence and local scope for exclusivity, including restore/reset, and retain the source's subject when stating applicability.
- Keep all actual0.1.8 failures and partial/unrun paths; require new installation and continuous model acceptance without changing frozen inputs or criteria. Refs #217, #220.
[0.1.8] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.8 keeps added display defaults separate from source facts, removes superseded clauses and exceptions from current evidence, and limits exclusivity to its stated operation scope.
- Make every reliable self-check premise explicit; source-supported feedback can remain determinate within a locally marked scenario assumption. Preserve decisive known AND/OR conditions in all prose and feedback.
- Keep the actual0.1.6 failures and the intermediate0.1.7 installation with model behavior NOT_RUN; verify the complete correction in fresh installed sessions. Refs #217, #220.
[0.1.7] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.7 separates initial/assigned values and reachable-state properties from a complete accepted-input domain. Source omissions remain explicit; added teaching constraints need local construction/design attribution.
- Keep known source-covered operations and field results despite unrelated input-domain gaps; an attribution elsewhere does not repair an unmarked restriction here.
- Preserve the actual0.1.6 fidelity failure and full native loading evidence; require new installation, affected paths and all six blocking groups. Refs #217, #220.
[0.1.6] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.6 compacts the entry skill after observed native prompt truncation on the baseline CLI; preserve full semantic methods in the existing two references.
- Enforce a conservative 6144-byte UTF-8 entry budget in repository and installed-resource validation, including byte-identical oversized packages.
- Preserve truncated/partial calls separately and require fresh cache, full native loading and behavior verification. Refs #217, #220.
[0.1.5] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.5 applies the text-only boundary to refusals and source quotations: summarize necessary meaning without echoing supplied HTML, code, pseudocode or production commands.
- Distinguish mutually exclusive operation branches from cumulative feedback; audit broad rows for overlap without losing known missing conditions or unknown-information notices.
- Preserve the actual 0.1.3 HTML-echo failure and repeated-operation failures. The intermediate 0.1.4 installation passed but its model behavior was not tested; verify the complete correction on a fresh canonical cache.
[0.1.4] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.4 rechecks source preconditions against the current state on repeated operations. Existing field results remain known when the rule still applies; unspecified extra feedback does not erase them.
- Preserve the actual 0.1.3 repeated-operation failure and invalid operator selections separately. Require new canonical installation, affected cases and all six blocking groups with actual target-heading selections.
[0.1.3] - 2026-10-13
Fixed
- HTML Guide Kit 0.1.3 distinguishes a single operation in its current state from a later multi-step path. An undefined path does not erase an explicit prohibition on the direct operation.
- Audit operation assertions across guide body, simulations, feedback and boundary statements before delivery. Preserve the actual 0.1.2 B4.1 failure separately from its successful B4.2 correction.
- Keep source terms and value types distinct from derived predicates; compressed headings and tables cannot add unsupported exclusive conditions.
[0.1.2] - 2026-10-12
Fixed
- HTML Guide Kit 0.1.2 preserves decisive known conditions when combining AND/OR with unknown information. Check branch overlap and priority without extending a local condition result into undefined actions or values.
- Preserve the actual 0.1.1 partial-rule simulation failure and require new installation, affected cases and all six blocking groups on this candidate.
[0.1.1] - 2026-10-12
Fixed
- HTML Guide Kit 0.1.1 cross-checks each operation's source preconditions and allowed states against narrative, demonstrations, unavailable feedback and self-checks. A correct rule elsewhere cannot cancel contradictory feedback.
- Preserve unspecified branches instead of deriving prohibition, opposite results or unchanged values from a rule that describes only one state. Use negative feedback only when the source establishes the corresponding necessary condition, exclusive scope or prohibition.
- Attribute added inspection axes and teaching suggestions to design judgment instead of describing them as explicit source rules or omissions. Preserve historical 0.1.0 failures and require new installation and behavior evidence for this correction candidate.
[0.1.0] - 2026-10-12
Added
- Prepare HTML Guide Kit 0.1.0 with the single explicit Read entry guide. Read supplied documents or a clear conversation snapshot, preserve source boundaries, offer distinct reader/task/understanding-question options and wait for explicit confirmation.
- Deliver copyable guide text in chat with depth adapted to the material. Describe faithful relationships, conditional demonstrations and retryable teaching feedback with reliable or bounded reference criteria.
- Preserve the new-guide-only scope and avoid HTML/code/file production, production-tool instructions and capability scores or long-term mastery claims. This version entry records a package draft; registration, client acceptance and publication require their own actual evidence.