Changelog

All notable changes to AutoMCM-Pro are documented here.
Format follows Keep a Changelog.
Version scheme: Semantic Versioning.

---

[0.3.1] — 2026-09-11

Added

• LeanGate optional Lean 4 addon with pinned toolchain, exact resource-allocation certificate generation, versioned claim manifest, semantic review/freeze workflow, kernel verification, forbidden-token audit, freshness invalidation, JSON reports, reproducible artifact export, and selected/required policies.
• quality_gate.py formal plus LeanGate checks in all and launch-check; pipeline_manager.py advance now refuses a verify-stage transition when an enabled formal gate fails.
• Mandatory per-problem formal-applicability classification before leaving problem_analysis, with immutable sealed baselines and automatic mandatory promotion for safety, feasibility, conservation, invariant, convergence, certificate, and decision-critical risk categories.
• LeanGate instructions for Claude Code, Codex, OpenCode, and DSH bindings.
• Receipt-backed lifecycle management for Claude Code, Codex, OpenCode, and DSH: selective install/uninstall, symlink or managed-copy mode, integrity checks, atomic copy updates with rollback, and conservative cleanup.
• Optional installation of the pinned Lean toolchain through Elan; uninstall only removes a toolchain when the receipt records that this installer created it.

Security

• LeanGate rejects sorry, admit, unsafe, custom axioms, path traversal, symlinked Lean inputs/directories (including .lake), unparseable axiom reports, stale approval evidence, and claim/obligation registry tampering.
• Certificate claims must name an exact JSON certificate, and generated Lean files embed its SHA-256 digest. Specs, proofs, certificates, project configuration, all local Lean sources, claim scope, and approval evidence participate in freshness.
• Export and launch gates independently rerun Lean compilation and #print axioms; copied reports or Agent-authored PASS text cannot satisfy the gate.
• Stage approvals are bound to stage and review round, preventing reuse of stale [APPROVED] markers.

Fixed

• Preserve installer ownership of Elan/toolchains across updates so an owned Lean toolchain remains safely removable without permitting deletion of user-managed ones.
• Refuse unsafe or malformed installation receipts, unmanaged Skill destinations, partial-install orphan links, and silent deletion of an unresolved rollback backup.
• Require epistemic_type=certificate claims to provide certificate_file; reject symlink directories before excluding ordinary Lean build caches.

Validation

• Added 31 automated tests covering LeanGate unit, end-to-end kernel, security, pipeline, and install/update/uninstall behavior.
• Verified Lean builds with the pinned leanprover/lean4:v4.32.0 toolchain.
• Independently reviewed and executed by OpenCode and DeepSeek Harness (DSH), both PASS.

Changed

• Renamed the formal-verification addon uniformly to LeanGate (leangate on disk and in CLI paths) while retaining AutoMCM-Pro as the product name.
• Added a bilingual project-classification, limited-purpose, data/privacy, and system-permission disclaimer clarifying that AutoMCM-Pro is a locally invoked coding-agent Skill rather than a model or independently operating agent.
• First interactive installation now asks whether to install the core Python modeling dependencies and whether to prepare LeanGate; non-interactive installation uses explicit --with-python-deps and --with-lean opt-ins.
• The four agent bindings now ask once on their first Skill invocation in each new session; they do not pre-scan or silently install dependencies, and non-interactive runtimes pause.
• The native Mind-Reader launcher now reports missing dependencies and exits unless the user explicitly authorizes installation with --install.
• setup_workspace.py and pipeline_manager.py init accept --lean-policy while preserving the previous default (off).
• Human stage approvals are round-scoped ([APPROVED:<stage>:<round>]) so stale approval markers cannot authorize later pipeline rounds.

[0.3.0] — 2026-08-25

Added

Los Alamos Exploration Layer (optional, auto-triggered addon)

• Hypothesis tree/DAG materialized view (scripts/los_alamos/hypothesis_tree.py), event-sourced from an append-only message bus (bus.py); four evaluation tracks: Track 0 (semantic screening), Track R "Bletchley" (red-team stress test), Track 1 (entropy-weight + TOPSIS objective ranking), Track 2 (qualitative pairwise judge panel, Copeland ranking, Shannon-entropy disagreement)
• Six autonomous trigger criteria (SKILL.md path C Step 0), including a new "problem-complexity initial judgment" checkable as early as problem_analysis — no longer requires an explicit human request or literature-survey-stage evidence to fire
• New LOS_ALAMOS_METHOD_CATALOG.md — six-category MCM paradigm taxonomy (optimization / prediction / evaluation / simulation / classification / network) used by Alsos literature surveys and Director's first-layer branch selection to widen candidate coverage beyond ad-hoc search
• Elastic branch width (2 branches under time pressure, 3–4 when time budget allows) and deeper branching support (same-paradigm technical-choice sub-branches, depth + --parent-ids)
• Governance rule (SOP §9.2.6): branch decisions must query hypothesis_tree.py status before acting, not rely on narrative memory — wired into SKILL.md at Step 5 (build/verify dispatch) and Step 9 (finals)
• Fixed: adjudicate.py screen/redteam previously left the hypothesis_tree_{N}.json materialized-view snapshot stale until the next hypothesis_tree.py invocation; both now call _sync_tree() immediately

RAND Delphi — Track 2 anonymous multi-round convergence

• adjudicate.py panel-vote --round / panel-entropy --round / delphi-summary; capped at 2 retry rounds (SOP §9.6)

Project Skunk Works — lightweight mode

• pipeline_manager.py init --skunk-works, relaxes citation gate (≥1 vs ≥2) and Checkpoint report verbosity; §4 self-verification mandate is explicitly not relaxed
• Autonomous trigger extended (SOP §10.1): Director may self-trigger based on a time-pressure signal alone (not only explicit human request), but must announce the switch before executing — never silent

Andon Cord — universal emergency stop

• pipeline_manager.py andon-pull/andon-clear/andon-status; freezes advance pipeline-wide until cleared; AP mode cannot auto-clear (one of only two hard human-in-the-loop exceptions, alongside Checkpoint LA)

NASA Mission Control Go/No-Go — final_compile pre-flight check

• quality_gate.py launch-check (gate 8): polls Andon / stage completeness (now filtered by problem_count, see Fixed) / image existence / placeholder residue / anonymity / Los Alamos assumption-ledger integrity
• Mandatory before final_compile (new absolute-prohibition item, SOP §7)

Project Kaizen — post-verification quality-polish loop

• pipeline_manager.py kaizen-assess/kaizen-round-start/kaizen-status; 4-dimension self-assessment (significance / robustness / assumptions / completeness, threshold 4.0/5), capped at 2 rounds, gated by the same objective time-budget signal as Skunk Works (opposite direction: adds rigor when time allows rather than relaxing it under pressure)
• Hard boundary vs. writing-style polish (§16): Kaizen may revise numeric conclusions and model content; style passes may not

Work Log — single-file complete session record

• scripts/worklog.py — append-only, Simplified Chinese, CUMCM_Workspace/memory/worklog.md; stage transitions / rework / Andon / Kaizen / gate results auto-logged from pipeline_manager.py and quality_gate.py at zero extra token cost; user messages and undocumented Agent decisions logged manually via worklog.py append

Citation Authenticity Verification & Shared Pool

• scripts/cite_check.py — register (dedup by DOI/URL/title into a shared citations.bib across sub-problems) / verify (DOI via CrossRef API, URL via HEAD request — catches hallucinated citations, not just format-shape matching) / list / export-bibitems
• quality_gate.py lit gate now prioritizes the verified citation pool over the legacy prose-regex scan; legacy path still works with a migration nudge
• Fixed: export-bibitems did not escape LaTeX special characters — a real DOI containing an underscore broke xelatex compilation with a fatal error

Writing-Style / "AI-Generated Feel" Reduction

• scripts/style_check.py scan — six objective, non-LLM checks: list-density in narrative sections, stock-phrase repetition, mechanical first/second/finally transition skeletons, sentence-length coefficient of variation, paragraph-length CV, cross-section n-gram repetition; thresholds informed by public research (mrshibly/Humanizer, Wikipedia:Signs of AI Writing, Pangram) and recalibrated against real Chinese academic writing
• SOP §16 writing protocol: plan-then-segment writing, ≤2 style-polish rounds; hard boundary — style polish may only change how something is said, never numeric conclusions (Kaizen's territory)
• Fixed: LaTeX comment lines (%% ==== section divider ====) were leaking into n-gram frequency analysis as false "repeated phrases"
• Fixed: overlapping sliding-window n-gram duplicates of the same longer phrase (e.g. three 6-char windows of one 8-char term) now merge into one finding instead of three

Official Format Compliance (SOP §17)

• CUMCM: templates/latex_template.tex synced to the 2026 format spec — electronic-version-compliant cover-page structure (title+abstract combined as page 1, page numbering starts there), \tableofcontents removed (official rule: no TOC), commented-out print-only 承诺书/编号页 block
• New anonymity heuristic gate quality_gate.py anon-check (self-referential / labeled-field phrase matching), wired into launch-check
• New scripts/ai_usage_doc.py: generate (builds a standalone 《AI工具使用详情》PDF from worklog.md per the 2025-trial AI-tool-usage regulation's four required sections), cite-format (CUMCM reference-line formatter), mcm-entry (COMAP Report on Use of AI Tools entry formatter) — none of these auto-insert into the paper; declaring AI usage is a deliberate, manual decision by design
• MCM/ICM: templates/mcm_template.tex gets \setcounter{tocdepth}{2} (keeps the required TOC to ~1 page) and a commented-out Report on Use of AI Tools placeholder section (COMAP format, appended after the 25-page solution, not counted toward the limit)
• compile_pdf.py now reports page counts after a successful compile (PDF-bookmark-based body/appendix split) with contest-specific soft targets and hard-cap warnings — CUMCM body target 20–25 pages / hard cap 30 (appendix unlimited); MCM/ICM hard cap 25 pages total including appendix code — and recommends compacting appendix core code first when over budget

§18 — Research domain-conventional visualization forms before drawing

• New AutoMCM_SOP.md §18 ("画图前先查领域惯例"): before designing any result figure, search (web-search, keywords abstracted per S3) for the conventional visualization form for that problem archetype — e.g. trajectory/geometry-optimization problems → 3D trajectory + key-moment annotation; coverage/siting/allocation → map/scatter + Gantt-style timeline; multi-objective/multi-scenario comparison → Pareto front or radar chart; sensitivity analysis → tornado chart or multi-parameter curve family; classification/clustering/ranking → heatmap or a re-sorted stacked bar chart. Findings recorded in thought_process.md; falling back to common-sense judgment is permitted when no clear convention is found, but — same spirit as §7's "never silently skip a failed check" — that fallback must also be recorded, never silently taken.
• Wired into .claude/skills/auto-mcm/SKILL.md's 图表风格规范 section (new subsection preceding 标准用法) and cross-referenced across all three lean runtime bindings (.opencode/, .dsh/, .agents/).
• Codex binding's existing "known capability gap: no native web search" section extended to explicitly cover this new requirement too — the same underlying gap (no WebSearch/WebFetch tool) blocks the domain-convention research step, not just literature review; the existing "disclose at Checkpoint① rather than silently skip" guidance now applies to both.

Two new fully-autonomous demo cases (replacing the previous single showcase)

• demo/2020C/ — real 2020 CUMCM Problem C (Credit Strategy for Small and Micro Enterprises), official attachments (~1.1M invoice rows) downloaded and verified from mcm.edu.cn. Full dsh headless run (zero human checkpoints) through all three sub-problems + sensitivity analysis: logistic-regression default-probability model (5-fold CV AUC 0.935), entropy-weight/TOPSIS cross-validation track, PSI transfer-stability check, KMeans industry-behavior clustering, expected-profit-maximizing credit allocation. 30 pages (21-page body), 18 figures, all 2D, zero 3D plots.
• demo/2025A/ — 2025 CUMCM Problem A (smoke-screen decoy deployment), restyled end-to-end into a NASA technical-report / orbital-mechanics chart theme (white background, dense fine grid + bold major gridlines, black/white primary lines, NASA Blue #0B3D91 / NASA Red #FC3D21 reserved for key-annotation accents only, fully boxed axes) — implemented as a workspace-local override of scripts/plot_style.py's palette/rcParams/ save() grid post-processing, not a change to the repository's shared style module. 28 pages (20-page body), same modeling conclusions as before restyling.
• demo/mind_reader_ui.png — the Mind-Reader tool screenshot (previously nested inside the single demo's CUMCM_Workspace/) relocated to a case-independent top-level path, since it illustrates the tool, not either specific paper.

Runtime Bindings

• New: DeepSeek Harness (dsh) binding — .dsh/skills/auto-mcm/SKILL.md
• New: opencode binding — .opencode/skills/auto-mcm/SKILL.md
• New: Codex CLI binding — .agents/skills/auto-mcm/SKILL.md
• All three verified via real executed tasks (not documentation inference), cross-checked through structured runtime session logs (opencode export, decompressed .zstd dsh session logs, plain-.jsonl Codex rollouts)

Fixed

• templates/latex_template.tex and templates/mcm_template.tex both failed to compile out of the box — both loaded \usepackage{natbib} while their bibliography section used plain \bibitem/\cite (not natbib's \citep/\citet), which is a fatal incompatibility (Package natbib Error: Bibliography not compatible with author-year citations). Neither template body actually uses any natbib-specific command, so natbib was unnecessary; removed from both. Verified with a from-scratch standalone xelatex compile of each template (CUMCM: exit 0, 6 pages, zero errors once a placeholder figure exists; MCM/ICM: confirmed the natbib error is gone, remaining failure is the separate, already- documented mcmthesis.cls external-package requirement). This bug predates this release — every fresh template compile would have hit it — and had only ever been patched ad hoc inside individual generated papers' main.tex, never at the template source; this is the first fix at the actual source. README prerequisites also updated to document the mcmthesis package requirement for the MCM/ICM template specifically (most minimal TeX Live installs don't ship it; CUMCM doesn't need it).
• plot_style.py: correlation-matrix heatmap used a one-directional sequential colormap for signed (−1…1) data instead of diverging_cmap() — negative and positive correlation were nearly indistinguishable at a glance; fixed.
• Two instances of a "single-x-tick stacked bar chart degenerates into a solid-color rectangle with no visible structure" bug, both triggered when an expected-profit-maximizing credit allocation concentrated almost entirely into one rating tier (2020C demo, problem1_credit.py and problem2_credit.py) — replaced with per-enterprise sorted horizontal bar charts (amount + PD annotated per bar), which is both correct rendering and more informative than the aggregate single bar it replaced.
• LaTeX verification-summary table (2025A demo main.tex) overflowed the page's right margin by ~415pt (~5.7in) — a plain lcl tabular with long cell text and no wrap points; fixed with array package p{}-width wrapped columns, plus \allowbreak inserted after underscores in monospace filenames (which otherwise have no natural line-break point and still overflow their column even when wrapping is enabled).
• 3D scene-overview and trajectory figures (2025A demo) had large dead whitespace bands where set_box_aspect'd axes were never tightened to their actual rendered content — added plot_style.fit3d_axes() (two-pass render + get_tightbbox()-based repositioning) and wired it into the affected 00_data_eda.py/problem1_geometry.py figure functions; also replaced two 3D-space text labels that sat close together in this scenario's real geometry (and so collided under certain view angles) with projection-immune screen-space text2D anchors.
• Documented, and deliberately reverted rather than shipped: a third 3D inset panel + taller figure canvas intermittently caused bbox_inches='tight''s save-time crop to drop unrelated existing labels from a sibling subplot on the same figure — reproduced twice independently (including after reducing label offsets), suspected to be an mplot3d Text3D bounding-box quirk rather than a usage error in this codebase; reverted to the previously-verified two-panel layout instead of risking a silent regression for a marginal enhancement.
• gate_launch_check() stage-completeness check now filters by problem_count — previously always scanned model_1–model_3 regardless of actual sub-problem count, producing a false permanent NO-GO for any 1- or 2-problem contest
• Hardcoded /tmp/ absolute paths in quality_gate.py example commands replaced with in-workspace paths — broke under sandboxed runtimes' external_directory permission checks

Testing

• Every bug above (degenerate allocation charts, dead 3D whitespace, table overflow, wrong colormap) was found by direct human/Claude visual inspection of the rendered PDF pages — none were caught by the dsh agent's own completion report, and none could have been: the model dsh was running (deepseek-v4-flash) has no image-input support, so its claimed "visual self-review" was necessarily only ever a file/pixel-level programmatic check, disclosed honestly in worklog.md rather than misrepresented.
• Multiple full real-agent pipeline runs (dsh, opencode) exercising the complete new addon set together, independently re-verified rather than trusting agent self-reports (PDF validity/page counts, launch-check re-run, cite_check verify re-run, session-log decompression for tool-call audit)
• Los Alamos genuinely self-triggered via multiple distinct criteria across runs (explicit "compare methods" problem wording, autonomous complexity judgment with no such wording); one run surfaced two genuine Checkpoint LA two-track conflicts (one directional contradiction, one Condorcet cycle) — resolved with fully documented headless-degradation reasoning; flagged as a known test-methodology limitation that Checkpoint LA's true human-wait path remains unverified in a genuinely interactive session

---

[0.2.0] — 2026-05-03

Added

/draw-image Skill — OpenAI gpt-image-2 图像生成

• New skill .claude/skills/draw-image/SKILL.md for generating flowcharts, architecture diagrams, and conceptual illustrations via OpenAI gpt-image-2.
• New script scripts/draw_image.py — CLI wrapper around the OpenAI Images API:
  • Supports gpt-image-2 (default), gpt-image-1.5, gpt-image-1, gpt-image-1-mini, dall-e-3
  • Parameters: --size (any valid WxH for gpt-image-2, up to 3840px), --quality (low/medium/high/auto), --output-format (png/jpeg/webp), --compression, --background, --moderation
  • --check flag: probes authentication without generating anything
• Three-tier auth detection (in priority order) with explicit default-on/off behavior:
  1. OPENAI_API_KEY env var → OpenAI Python SDK path (token-based billing)
  2. Codex OAuth session detected (scans ~/.codex/auth.json and related paths, CODEX_AUTH_TOKEN / OPENAI_OAUTH_TOKEN env vars) → auto-enabled without any configuration; routes to Codex CLI $imagegen (free within ChatGPT Plus/Pro subscription)
  3. Neither configured → disabled by default; graceful skip (exit code 2, not an error); pipeline continues with \missingfigure{} placeholder in LaTeX
• Design intent: Codex users get image generation out of the box; non-Codex users see the feature as off unless they explicitly supply an API key — no unexpected charges, no pipeline interruptions
• Skill documents: decision tree (when to use AI images vs. matplotlib), prompt engineering templates, LaTeX integration, pricing table, error guide

Multi-Agent Parallel Pipeline (AP Mode)

• pipeline_manager.py new flag and commands:
  • init --problems N — records sub-problem count; enables automatic AP multi-agent parallelism when N > 1; stored as problem_count in pipeline.json
  • suggest-parallel — inspects current pipeline state and outputs the next batch of stages that can be parallelized right now (exit 0 + space-separated stage list); exit 1 when nothing to parallelize (single problem or conditions not met). Two phases auto-detected:
    1. data_preprocessing approved → outputs all model_N_build stages
    2. All builds approved → outputs all model_N_verify stages
  • parallel-start <s1> <s2> ... — mark multiple stages in_progress simultaneously, enabling concurrent Agent execution
  • parallel-status <s1> <s2> ... — print a completion table for a stage group
  • parallel-all-done <s1> <s2> ... — exit 0 if all stages are approved, exit 1 otherwise (safe to use in shell conditionals)
• pipeline_manager.py new constant PARALLEL_GROUPS documenting which stages are safe to parallelize and their prerequisites
• auto-mcm/SKILL.md AP pipeline section updated:
  • model_build + model_verify entry now has a parallel decision gateway: calls suggest-parallel first; Path A (parallel) or Path B (sequential)
  • AP sub-Agent prompt template included inline with AP-mode self-approve step
  • 【多 Agent 并行策略】 section refactored into a concise reference with command quick-reference table; full steps moved to AP flow section
  • Background draw-image dispatch pattern with graceful skip
  • LaTeX section parallelization via latex/sections/ fragments

Contest Git — 竞赛工作区版本控制

• New script scripts/contest_git.py — manages an independent Git repo inside CUMCM_Workspace/ (separate from the AutoMCM-Pro tool repo):
  • init — creates the workspace repo with .gitignore (LaTeX intermediates excluded) and an initial commit; configures git identity automatically
  • auto_commit(stage, mode, round_n) — called by pipeline_manager.py after each advance; commits staged changes with semantic message feat(<stage>): approved [AP] or fix(<stage>): rework rN approved
  • rework_start(stage, round_n) — empty commit marking the rework entry point in the log, called on rework
  • milestone_tag(name, message) — annotated tag (overwrites if exists)
  • Auto-tagging: latex_draft approved → draft-v1, final_compile approved → final-v1 (increments with review rounds)
  • Read-only queries: log(n, oneline), diff(ref1, ref2, stat_only), status(), list_tags()
  • Standalone CLI: python scripts/contest_git.py init|log|diff|status|tag|tags
• pipeline_manager.py integration:
  • init --git flag: enables contest git and calls contest_git.init()
  • git_enabled field persisted in pipeline.json
  • advance auto-commits on every stage approval
  • rework records a rework-start empty commit
  • New contest-git subcommand group: log, diff, status, tag, tags
• auto-mcm/SKILL.md: new 【竞赛工作区版本控制】 section with event/git-action table and usage examples

Skill Updates

• cumcm-master/SKILL.md: figure source decision tree (data vs. non-data); /draw-image integration in the figure-generation step
• mcm-master/SKILL.md: same figure decision tree in English
• auto-mcm/SKILL.md: latex_draft stage now documents draw-image option

Changed

Security Fixes

• pipeline_manager.py: sanitize user-provided --summary/--results/ --concerns before writing to eval log and review files — prevents injection of [APPROVED]/[REWORK] control markers via argument strings
• pipeline_manager.py: marker replacement in human_intervention.md now uses str.replace(..., count=1) — preserves full review history; prior approvals and reworks no longer silently overwritten
• pipeline_manager.py: load() now catches json.JSONDecodeError and exits with a clear recovery message instead of an uncaught exception
• pipeline_manager.py: emit UserWarning when contest_git module is unavailable instead of silently disabling all version-control features
• contest_git.py: auto_commit now calls _scan_staged_for_secrets() before every commit; blocks and warns if OpenAI keys (sk-…), GitHub tokens (ghp_…), AWS keys (AKIA…), or password=/token= patterns are found in staged diffs
• contest_git.py: .gitignore expanded to block .env, *.key, *.pem, credentials.*, secrets.*, *_token*, config.local.*

Automation & Reliability

• pipeline_manager.py: --max-reworks N flag on init (default 5); cmd_rework checks per-stage rework count and exits with code 2 if limit exceeded, preventing infinite repair loops
• auto-mcm/SKILL.md UX overhaul — zero command-line interaction for users:
  • Wake-up protocol now checks initialization state first; triggers 首次启动协议 (first-launch protocol) when workspace is not yet set up
  • First-launch protocol: agent asks for file paths via natural language (AskUserQuestion), reads the problem to auto-detect sub-problem count and contest type, then runs all init commands silently (setup_workspace.py + pipeline_manager.py init with --problems N --git)
  • AP mode checkpoints: replaced raw command output with natural language progress reports after each stage; no user input needed
  • MANUAL mode: user provides specs and approvals in natural language; agent translates internally to pipeline commands; removed "输入继续" terminal instruction
  • Rework protocol: user states changes in natural language; agent writes to human_intervention.md and executes rework command silently
• README.md "How to Use" section rewritten for zero-command UX:
  • Removed all pipeline_manager.py init instructions from user-facing steps
  • Added AP mode and MANUAL mode natural language dialogue examples
  • Updated English section to match
• New script scripts/quality_gate.py — hard-enforced 4-gate quality CLI (not behavioral rules): literature reference count (≥2 per sub-problem), numerical sanity scan (inf/nan/1e200+), structured PASS/FAIL report parsing, physical-constant cross-problem consistency check; subcommands verify / sanity / lit / consist / all; exit codes 0=pass, 1=fail, 2=skip
• New script scripts/security_check.py — hard-enforced security CLI: path traversal prevention (check_paths), env-variable leak detection (check_env_not_leaked), workspace secret scan (scan_files_for_secrets, scan_workspace_all); 6 secret patterns (OpenAI, Anthropic, GitHub, AWS, password literals, generic API keys); output auto-redacted to ≤20 chars
• auto-mcm/SKILL.md: 【建模质量门控】 updated — now documents exact python scripts/quality_gate.py invocations for every pipeline moment (build pre-gate, build post-gate, verify gate, consistency gate); structured ===VERIFICATION REPORT=== contract; LaTeX 3-attempt retry loop
• auto-mcm/SKILL.md: new 【安全规程】 section — API key protection, file path validation via security_check.py path, external-service query abstraction, rework-limit user notification, secret-commit interception
• auto-mcm/SKILL.md: dependency auto-check step added to wake-up protocol
• pipeline_manager.py docstring updated to list all new commands

Removed

• demo/multi_agent_demo/ — development-only test case (verified and removed after successful end-to-end run: 22/22 verifications passed, PDF compiled, gpt-image-2 flowchart generated)

---

[0.1.0-beta] — 2026-03-14

Added

• Initial release of AutoMCM-Pro
• /auto-mcm, /cumcm-master, /mcm-master skills
• AP / MANUAL dual-mode pipeline
• GitOps state machine (pipeline_manager.py) with 8 commands
• Mandatory self-verification protocol (verify_*.py for every solver)
• Mind-Reader real-time thought visualization (FastAPI + WebSocket)
• Docker environment (Python + TeX Live)
• CUMCM and MCM/ICM LaTeX templates
• Demo: CUMCM 2025 Problem A (11 stages, 144 verifications, ~1h34m runtime)