Add bloom-tutor skill: interactive Socratic learning system#76
Open
Li-Evan wants to merge 158 commits into
Open
Add bloom-tutor skill: interactive Socratic learning system#76Li-Evan wants to merge 158 commits into
Li-Evan wants to merge 158 commits into
Conversation
…ents ## New Skill: continue-claude-work (v1.1.0) - Recover actionable context from local `.claude` session artifacts - Compact-boundary-aware extraction (reads Claude's own compaction summaries) - Subagent workflow recovery (reports completed vs interrupted subagents) - Session end reason detection (clean exit, interrupted, error cascade, abandoned) - Size-adaptive strategy for small/large sessions - Noise filtering (skips 37-53% of session lines) - Self-session exclusion, stale index fallback, MEMORY.md integration - Bundled Python script (no external dependencies) - Security scan passed, argument-hint added ## Skill Updates - **skill-creator** (v1.5.0): Complete rewrite with evaluation framework - Added agents/ (analyzer, comparator, grader) - Added eval-viewer/ (generate_review.py, viewer.html) - Added scripts/ (run_eval, aggregate_benchmark, improve_description, run_loop) - Added references/schemas.md (eval/benchmark schemas) - Expanded SKILL.md with inline vs fork guidance, progressive disclosure patterns - Enhanced package_skill.py and quick_validate.py - **transcript-fixer** (v1.2.0): CLI improvements and test coverage - Enhanced argument_parser.py and commands.py - Added correction_service.py improvements - Added test_correction_service.py - **tunnel-doctor** (v1.4.0): Quick diagnostic script - Added scripts/quick_diagnose.py - Enhanced SKILL.md with 5-layer conflict model - **pdf-creator** (v1.1.0): Auto DYLD_LIBRARY_PATH + rendering fixes - Auto-detect and set DYLD_LIBRARY_PATH for weasyprint - Fixed list rendering and CSS improvements - **github-contributor** (v1.0.3): Enhanced project evaluation - Added evidence-loop, redaction, and merge-ready PR guidance ## Documentation - Updated marketplace.json (v1.38.0, 42 skills) - Updated CHANGELOG.md with v1.38.0 entry - Updated CLAUDE.md (skill count, marketplace version, daymade#42 description) - Updated README.md (badges, skill section daymade#42, use case, requirements) - Updated README.zh-CN.md (badges, skill section daymade#42, use case, requirements) - Fixed absolute paths in continue-claude-work/references/file_structure.md ## Validation - All skills passed quick_validate.py - continue-claude-work passed security_scan.py - marketplace.json validated (valid JSON) - Cross-checked version consistency across all docs
- Add _ensure_list_spacing() to handle lists without blank lines before them - Modify _md_to_html() to preprocess markdown content via stdin - Add automated test suite (scripts/tests/test_list_rendering.py) - Fix: Lists without preceding blank lines now render correctly - Original markdown files remain unmodified (preprocessing in memory only) Root cause: Pandoc requires blank lines before lists per CommonMark spec. Without preprocessing, lists following paragraphs render as plain text. Tested scenarios: ✅ Lists with blank lines (normal case) ✅ Lists without blank lines (critical fix) ✅ Ordered lists without blank lines ✅ Original file integrity preserved Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…on workflow - Add 10-point high-quality PR formula based on real-world success cases - Add investigation phase workflow (post to issue before PR) - Add git history tracing techniques (git log, git blame) - Add evidence-loop pattern (reproduce → trace → link → post) - Add high-quality PR case study reference - Update PR checklist with investigation steps - Emphasize separation of concerns (detailed analysis in issue, fix summary in PR) Key principles: - Deep investigation before coding - Minimal, surgical fixes - Professional communication - No internal/irrelevant details in PR Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ripts New scripts: - fix_transcript_timestamps.py: Repair malformed timestamps (HH:MM:SS format) - split_transcript_sections.py: Split transcript by keywords and rebase timestamps - Automated tests for both scripts Features: - Timestamp validation and repair (handle missing colons, invalid ranges) - Section splitting with custom names - Rebase timestamps to 00:00:00 for each section - Preserve speaker format and content integrity - In-place editing with backup Documentation updates: - Add usage examples to SKILL.md - Clarify dictionary iteration workflow (save stable patterns only) - Update workflow guides with new script references - Add script parameter documentation Use cases: - Fix ASR output with broken timestamps - Split long meetings into focused sections - Prepare sections for independent processing Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add skill to fix broken line wrapping in Claude Code exported .txt files. Reconstructs tables, paragraphs, paths, and tool calls that were hard-wrapped at fixed column widths. Features: - State-machine parser with next-line look-ahead - Handles 10 content types (user prompts, Claude responses, tables, tool calls, etc.) - Pangu spacing for CJK/ASCII mixed text - 53 automated validation checks - Safety: never modifies original files, verifies marker counts Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds dependency detection before skill creation starts, preventing mid-workflow failures (e.g., gitleaks missing at packaging, PyYAML missing at validation). Documents correct script invocation via python3 -m syntax and auto-installation commands. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…losure Move verbose sections to references/ files, keeping concise pointers in CLAUDE.md. Zero content loss — all documentation preserved in reference files that Claude loads on demand. Moved to references/: - plugin-architecture.md (296 lines) — architecture docs - plugin-troubleshooting.md (441 lines) — installation debugging - new-skill-guide.md (241 lines) — detailed templates/checklists - promotion-policy.md (60 lines) — third-party request policy - youtube-downloader/references/internal-sop.md — yt-dlp SOP Also fixed: Available Skills daymade#36-42 indentation, deduplicated 4x versioning sections into one, removed stale notes. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- add scrapling-skill with validated CLI workflow, diagnostics, packaging, and docs integration - fix skill-creator package_skill.py so direct script invocation works from repo root - fix continue-claude-work extract_resume_context.py typing compatibility for local python3 - bump marketplace to 1.39.0 and updated skill versions
Replace real Zhipu GLM API key with fake placeholder in mask_secret() and SecretStr docstring examples. The real key was exposed in this PUBLIC repo. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ndings transcript-fixer: - Add common_words.py safety system (blocks common Chinese words from dictionary) - Add --audit command to scan existing dictionary for risky rules - Add --force flag to override safety checks explicitly - Fix substring corruption (产线数据→产线束据, 现金流→现现金流) - Unified position-aware replacement with _already_corrected() check - 69 tests covering all production false positive scenarios tunnel-doctor: - Add Step 5A: Tailscale SSH proxy silent failure on WSL - Add Step 5B: App Store vs Standalone Tailscale on macOS - Add Go net/http NO_PROXY CIDR incompatibility warning - Add utun interface identification (MTU 1280=Tailscale, 4064=Shadowrocket) - Fix "Four→Five Conflict Layers" inconsistency in reference doc - Add complete working Shadowrocket config reference Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude IS the AI — when running inside Claude Code, use Claude's own language understanding for Stage 2 corrections instead of calling an external API. No API key needed by default. New capabilities in native mode: - Intelligent paragraph breaks at logical topic transitions - Filler word reduction (excessive repetition removal) - Interactive review with confidence-level tables - Context-aware judgment using full document context API mode (GLM) remains available for batch/automation use cases. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…AskUserQuestion New skill: asr-transcribe-to-text (v1.0.0) - Transcribe audio/video via configurable ASR endpoint (Qwen3-ASR default) - Persistent config in CLAUDE_PLUGIN_DATA (endpoint, model, proxy bypass) - Single-request-first strategy (empirically proven: 55min in one request) - Fallback overlap-merge script for very long audio (18min chunks, 2min overlap) - AskUserQuestion at config init, health check failure, and output verification skill-creator optimization (v1.5.1 → v1.6.0) - Add AskUserQuestion best practices section (Re-ground/Simplify/Recommend/Options) - Inject structured decision points at 8 key workflow stages - Inspired by gstack's atomic question pattern Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Rename skill to better reflect its purpose (document-to-markdown conversion) - Update SKILL.md name, description, and trigger keywords - Add benchmark reference (2026-03-22) - Update marketplace.json entry (name, skills path, version 2.0.0) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…gnosis Real-world findings from debugging docker build failures on macOS with OrbStack + Shadowrocket: - Add docker pull vs docker build vs docker run proxy path distinction table - Add 2G-1: --network host workaround for OrbStack transparent proxy broken by TUN - Rewrite 2G-2: use host.internal (not 127.0.0.1) for OrbStack Docker proxy - Add 2G-4: container healthcheck failure from lowercase http_proxy env var leak - Add 3 new symptom entries to Step 1 diagnostic index - Add smoking gun diagnosis: wget showing "127.0.0.1: Connection refused" Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… full rename cleanup - Add CJK bold spacing fix: insert spaces around **bold** spans containing CJK characters for correct rendering (handles emoji adjacency, already-spaced) - Add JSON pretty-print: auto-format JSON code blocks with 2-space indent - Add 31 unit tests covering all post-processing functions - Fix pandoc simple table detection (1-space column gaps) - Fix image path double-nesting when --assets-dir ends with 'media' - Rename all markdown-tools references across 15 files (README, QUICKSTART, marketplace.json, CLAUDE.md, meeting-minutes-taker, GitHub templates) - Add 5-tool benchmark report (Docling/MarkItDown/Pandoc/Mammoth/ours) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… to 2.1.0 Sync description with actual capabilities: CJK bold spacing, JSON pretty-print, simple table support, 31 tests, benchmark score. Add cjk/chinese keywords. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…kill draft - pdf-creator v1.2.0: theme system (default/warm-terra), dual backend (weasyprint/chrome auto-detect), argparse CLI, extracted CSS to themes/ - terraform-skill: operational traps from real deployments (provisioner timing, DNS duplication, multi-env isolation, pre-deploy validation) - asr-transcribe-to-text: add security scan marker Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…iew Team - Correct source accessibility: distinguish circular verification (forbidden) from exclusive information advantage (encouraged) - Add Counter-Review Team with 5 specialized agents (claim-validator, source-diversity-checker, recency-validator, contradiction-finder, counter-review-coordinator) - Add Enterprise Research Mode: 6-dimension data collection framework with SWOT, competitive barrier, and risk matrix analysis - Update version to 2.4.0 - Add comprehensive reference docs: - source_accessibility_policy.md - V6_1_improvements.md - counter_review_team_guide.md - enterprise_analysis_frameworks.md - enterprise_quality_checklist.md - enterprise_research_methodology.md - quality_gates.md - report_template_v6.md - research_notes_format.md - subagent_prompt.md Based on "深度推理" case study methodology lessons learned. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ample Replace \"深度推理（上海）科技有限公司\" with \"字节跳动子公司\" as the case study example to avoid exposing user's own company info. Also update .gitignore to exclude: - deep-research-output/ (contains sensitive research data) - recovered_deep_research/ - .opencli/ - douban-skill/ (work-in-progress) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…dology New skill: douban-skill - Full export of Douban (豆瓣) book/movie/music/game collections via Frodo API - RSS incremental sync for daily updates - Python stdlib only, zero dependencies, cross-platform (macOS/Windows/Linux) - Documented 7 failed approaches (PoW anti-scraping) and why Frodo API is the only working solution - Pre-flight user validation, KeyboardInterrupt handling, pagination bug fix skill-creator enhancements: - Add development methodology reference (8-phase process with prior art research, counter review, and real failure case studies) - Sync upstream changes: improve_description.py now uses `claude -p` instead of Anthropic SDK (no ANTHROPIC_API_KEY needed), remove stale "extended thinking" ref - Add "Updating an existing skill" guidance to Claude.ai and Cowork sections - Restore test case heuristic guidance for objective vs subjective skills README updates: - Document fork advantages vs upstream with quality comparison table (65 vs 42) - Bilingual (EN + ZH-CN) with consistent content Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…UDE.md) Prevents sensitive data (user paths, phone numbers, personal IDs) from entering git history. Born from redacting 6 historical commits. - .gitleaks.toml: custom rules for absolute paths, phone numbers, usernames - .githooks/pre-commit: dual-layer scan (gitleaks + regex fallback) - CLAUDE.md: updated Privacy section documenting the defense system Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…nciple, add date handling rules
- SKILL.md length driven by information density, not line count
- Factual dates (release dates, "last verified") should be kept — they help readers judge freshness
- Conditional date logic ("before X use old API") should be avoided
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… recovery Previously, recover_content.py saved all files flat in the output directory, causing files with the same name (e.g., src/utils.py and tests/utils.py) to overwrite each other. Now the script preserves the original directory structure, creating subdirectories as needed within the output directory. - Bump version: 1.0.0 → 1.0.1 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace hardcoded user paths that triggered gitleaks PII detection: - /Users/username/ → ~/ - /Users/user/ → ~/ - -Users-username- → -Users-<username>- (normalized paths) Also fix the sed example to use <home> placeholder instead of regex pattern that would match actual usernames. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Update SKILL.md and workflow_examples.md to reflect the new behavior of recover_content.py which now preserves original directory structure: - SKILL.md: Add 'preserving the original directory structure' note - SKILL.md: Update verification examples to use find command and show subdirectory paths (e.g., ./recovered_content/src/components/) - workflow_examples.md: Update diff example to account for nested paths Version bump: 1.0.1 → 1.0.2 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
asr-transcribe-to-text: - Add local MLX transcription path (macOS Apple Silicon, 15-27x realtime) - Add bundled script transcribe_local_mlx.py with max_tokens=200000 - Add local_mlx_guide.md with benchmarks and truncation trap docs - Auto-detect platform and recommend local vs remote mode - Fix audio extraction format (MP3 → WAV 16kHz mono PCM) - Add Step 5: recommend transcript-fixer after transcription transcript-fixer: - Optimize SKILL.md from 289 → 153 lines (best practices compliance) - Move FALSE_POSITIVE_RISKS (40 lines) to references/false_positive_guide.md - Move Example Session to references/example_session.md - Improve description for better triggering (226 → 580 chars) - Add handoff to meeting-minutes-taker skill-creator: - Add "Pipeline Handoff" pattern to Skill Writing Guide - Add pipeline check reminder in Step 4 (Edit the Skill) Pipeline handoffs added to 8 skills forming 6 chains: - youtube-downloader → asr-transcribe-to-text → transcript-fixer → meeting-minutes-taker → pdf/ppt-creator - deep-research → fact-checker → pdf/ppt-creator - doc-to-markdown → docs-cleaner / fact-checker - claude-code-history-files-finder → continue-claude-work Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
SKILL.md: rewritten following Anthropic best practices - Concise (233 lines, down from 347) - Critical VHS parser limitations section (base64 workaround) - Advanced patterns: self-bootstrap, output filtering, frame verification - Better description for skill triggering New files: - references/advanced_patterns.md: production patterns from dbskill project - assets/templates/self-bootstrap.tape: self-cleaning demo template auto_generate_demo.py: new flags - --bootstrap: hidden setup commands (self-cleaning state) - --filter: regex pattern to filter noisy output - --speed: post-processing speed multiplier (gifsicle) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Update 11 files to reflect the new plugin configuration pattern: - source points directly to skill directory (not repo root) - skills field omitted for single-skill plugins (auto-discovery) - skills field retained only for suite plugins (multi-skill routing) Files updated: - CLAUDE.md: marketplace config description, plugin count 57→55 - references/new-skill-guide.md: plugin entry template - references/plugin-architecture.md: config examples, install flow - references/plugin-troubleshooting.md: naming mismatch examples - marketplace-dev/SKILL.md: source/skills guidance, checklist, template - marketplace-dev/references/: cache patterns, anti-patterns, schema - skill-creator/SKILL.md: marketplace registration template - claude-skills-troubleshooting/references/architecture.md: schema example - marketplace-dev/hooks/post_edit_sync_check.sh: match skills via source path when skills array is absent Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bundle 5 audio/speech skills into a new suite covering the full pipeline:
recording → ASR transcription → transcript correction → meeting minutes → TTS
Suite members:
- asr-transcribe-to-text (from repo root)
- stepfun-asr (from repo root, newly tracked)
- transcript-fixer (from repo root)
- meeting-minutes-taker (from daymade-docs — semantic analysis,
not document format processing)
- stepfun-tts (from repo root)
Marketplace 1.53.2 → 1.54.0, plugin entries 55 → 56
(4 suites: daymade-audio, daymade-claude-code, daymade-docs, daymade-skill).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sync badge counts (skills 51→52, version 1.51.0→1.52.0) and add stepfun-asr skill entry in both English and Chinese READMEs. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Overhaul statusline-generator with improved context window display, health check script, customization guide, and troubleshooting decision tree. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- tunnel-doctor: add DNS resolver chain stall documentation (layer 5) - feishu-doc-scraper: track new skill files - Remove COMPETITOR_MATRIX.md (stale, not maintained) - Remove wechat-article-scraper path patterns from .pii-path-patterns (skill was previously unregistered from marketplace) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…minimal Most OpenAPI accounts cannot access application/skills-backend/* (gateway returns 1009 even with a fresh OAuth token), making the previous default --preset full a footgun: 16 of 19 skills error on every call. The 3 legacy skills (gangtise-data, gangtise-file, gangtise-kb) use public open-* endpoints and work on every account that can authenticate. - PRESET="minimal" is now the default in install_gangtise.sh - PRESET_WORKSHOP="$PRESET_MINIMAL" — workshop is an alias to stop footgunning live demos with the historical -client-heavy bundle - ISSUE-007 expanded with re-verification, 1009 vs 1008 disambiguation, 3-step diagnostic curl recipe, and warnings about workflow skill cascading failures + diagnose.sh false-positive greens - Sanitize references for public reuse (placeholders instead of project-specific examples) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…v1.5.0) Add regression tests for the CJK typography contract that has shipped since v1.4.0 but lacked test coverage. Document the overflow-wrap:normal trade-off and clarify the colgroup neutralizer's scope to default.css. No behavior change — implementation untouched. What's locked in: - _TYPOGRAPHY_CSS_PATCH: table-layout fixed, word-break keep-all, overflow-wrap normal (the trade-off "let content overflow slightly rather than fall back to mid-token breaks"), line-break strict, th nowrap - default.css colgroup neutralizer (default theme only; warm-terra and mobile use different strategies) - _lint_pdf_typography Layer 2 detector for 4 CJK anti-patterns New tests (16 total, all pass): - scripts/tests/test_cjk_tables.py — 7 contract + 2 end-to-end smoke - scripts/tests/test_typography_lint.py — 4 anti-pattern + 1 long-line exemption + 1 snippet capture + 1 negative control SKILL.md updates: - description: explicit scope boundary "markdown → PDF only; docx via minimax-docx" to prevent confusion in mixed workflows - Layer 1 patch description: document the overflow-wrap:normal trade-off and reference the locking test - colgroup neutralizer scope: clarify default.css only Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…-failure records
- Add SSR HTTP extraction fallback via browser_cookie3 + requests
when AppleScript/JXA/CDP are unavailable
- New script: scripts/download_feishu_images.py for standalone image
extraction with batch mode support
- Enforce per-document image naming ({doc-name}-{index}.ext) across
feishu_dom_capture.js and all docs
- Add Rule 17-19 to history-derived-rules.md covering SSR extraction,
per-doc naming, and [图片: Feishu Docs - Image] placeholder handling
- Record known failure paths (AppleScript disabled, JXA Promise errors,
CDP 404) in tooling-matrix.md with immediate fallbacks
- Update CLAUDE.md skill index to reflect new capabilities
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…1.6.0) The cjk-auto.css theme existed only in ~/.claude/plugins/cache/ and was lost on every skill upgrade. Copy it into version control so it ships with the skill. cjk-auto is a default.css derivative that uses table-layout: auto instead of fixed, letting column widths adapt to actual cell content. Best for tables with highly uneven column lengths (course schedules, itemized lists) where fixed equal-width forces CJK mid-breaks. Also update SKILL.md theme table, README/CHANGELOG, and bump marketplace to v1.55.0. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…1 + 3 sibling skills + gitleaks FP fix claude-md-progressive-disclosurer 1.2.1→1.3.1: integrate world-class context-engineering methodology (signal/anti-signal triage, ✅/⚠️ /🚫 tri-state vs priority inflation, @import context-cost warning, scope-misplacement hard-check, Why-per-rule, negative+positive pairing); self-fixed ≤500-line dogfood via verbatim relocation to refs A/B/C; eval-driven fix of R4 regression (mixed rule+case-study paragraph must move verbatim, refs 案例14). Eval: pass-rate 0.42→1.00 vs prior version, 7/7 discriminating assertions. Parallel-agent sibling skills (content updates, MINOR bumps): feishu-doc-scraper 1.1.0→1.2.0, gangtise-copilot 1.1.0→1.2.0, github-contributor 1.0.3→1.1.0. Throwaway feishu-doc-scraper-workspace/ deliberately excluded (not a skill). .gitleaks.toml: add \b word boundaries to phone-number-cn regex — 13-digit epoch-ms timestamps no longer false-match as CN mobile numbers (mirrors documented lib.sh patch). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
weasyprint subset-embeds PingFang SC as CID Type 0C OpenType, which macOS Preview and Adobe Reader fail to render — garbled text on recipient devices even though Chrome's PDF viewer shows it correctly. _detect_backend() now reads the markdown source: if CJK characters are found, auto-selects Chrome; non-CJK content still uses weasyprint. No user action needed — the fix is transparent. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…le residue Problem: 2-char rule 克劳→Claude matched before 3-char rule 克劳德→Claude, causing 克劳德 to become Claude德 (residue). Fix: ORDER BY from_text → ORDER BY LENGTH(from_text) DESC, from_text in all 4 get_all_corrections() query paths. Long rules now match first. Also fixed in local corrections.db (not in git): - 冈底斯→岗底斯 → 冈底斯→Gangtise (was mapping to wrong variant) - Deleted S10→S10 (tier ranking) (corrupted S101 → S10 (tier ranking)1) - Added 克劳德→Claude (covers short rule) - Added Claude扣→Claude (ASR variant) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Remove 17 standalone plugin entries from marketplace.json so suite member skills are reachable only via their suite namespace. Marketplace plugin entries: 56 → 39; version v1.55.0 → v1.56.0. This unifies daymade-audio, daymade-claude-code, and daymade-docs with daymade-skill (already suite-only). Each skill keeps its SKILL.md, version, and bundled scripts unchanged on disk under <suite>/<skill>/. Affected skills (17): - daymade-audio (5): asr-transcribe-to-text, stepfun-asr, stepfun-tts, transcript-fixer, meeting-minutes-taker - daymade-claude-code (7): claude-code-history-files-finder, continue-claude-work, claude-skills-troubleshooting, claude-md-progressive-disclosurer, statusline-generator, claude-export-txt-better, marketplace-dev - daymade-docs (5): doc-to-markdown, mermaid-tools, pdf-creator, ppt-creator, docs-cleaner Migration: run 'claude plugin marketplace update daymade-skills' then install the corresponding suite (e.g., 'claude plugin install daymade-audio@daymade-skills'). Invocation paths change from <skill>:<skill> → <suite>:<skill>. User data is safe — skills that persist data write to \$HOME (e.g., ~/.transcript-fixer/corrections.db). Cross-reference fixes (12 bare slash commands → namespaced): - deep-research, fact-checker, youtube-downloader, feishu-doc-scraper - daymade-docs/doc-to-markdown, daymade-claude-code/claude-code-history-files-finder - daymade-audio/transcript-fixer, daymade-audio/meeting-minutes-taker, daymade-audio/asr-transcribe-to-text Stale leftovers from v1.54.0 suite migration also cleaned up: - /daymade-docs:meeting-minutes-taker listing (moved to daymade-audio) - ./daymade-docs/meeting-minutes-taker/SKILL.md broken link - ./transcript-fixer/references/... broken links (orphan path) - 'Current: v1.53.0' stale version reference in CLAUDE.md (derived value) 15 README skill sections get a new install notice blockquote so users who land on a skill section know which suite installs it. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…olumn guidance (daymade#72) * docs(transcript-fixer): clarify native correction protocol + fix DB column guidance Make the AI-pass (Phase 2) correction flow explicit about behaviors that matter on long, hard transcripts, add guardrails for nested/weaker execution, and fix two doc-accuracy issues found in real use. No script/behavior changes. - Native flow: three-bucket triage (confident / needs-verification / uncertain), verify proper nouns by search instead of guessing, keep unconfirmable garbles as-is + emit a needs-checking list, second-pass review (with a fallback to a manual re-read when Task is unavailable, e.g. nested subagent context) - Scale-rigor escape hatch: short/clean transcripts skip the heavy machinery - Calibrate Stage 1 dictionary positioning (low hit-rate on high-quality ASR / specialized domains is expected; the AI pass does the real work) - Database Operations: real column names from_text/to_text (not wrong_term/ correct_term) + example queries -- guessing columns was the top failure mode - Quick Start points to the full method instead of duplicating a stale recap - Generalize examples (Whisper/Otter; medical/podcast/earnings) for a global audience Validated with skill-creator evals: on short/clean synthetic transcripts new and old flows are byte-identical (a strong model already handles those -- zero regression). On a long real transcript the old single-pass flow fabricated a product name out of a real one and missed several recurring errors that the new flow's triage + verify + second-pass caught. Net: no change on easy inputs, measurable gain on long/hard ones. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(transcript-fixer): scrub private real names + a real transcript snippet from examples The decision matrix, a domain example, and a CLI usage example used real personal / project names plus a verbatim line lifted from the author's own transcript. Swap them for public, generic equivalents (LangChain / Hugging Face, Karpathy / Anthropic, affect / effect, legal / gaming, and a placeholder section marker) that carry the same teaching point without exposing private content in a public repo. The transcript snippet was found by a full AI re-read of the file, not keyword scanning -- it had no name to grep for. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(transcript-fixer): bump daymade-audio to 1.1.0 MANDATORY version bump for the transcript-fixer SKILL.md content change in this PR (per CLAUDE.md "Updating Existing Skills"). Was missing from the original commits. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…tion method (daymade#73) * docs(skill-creator): make AI semantic read-through the primary sanitization method Scanners (gitleaks, grep, security_scan.py) only match patterns you thought to list -- they are structurally blind to private content with no keyword: a real name in another language, a verbatim line from a real transcript, a real example dropped into an illustration. A real leak slipped past keyword scanning exactly this way. - sanitization_checklist: lead with "read it yourself, don't just grep"; grep is a first pass, not the gate; "no matches" is not a clean bill of health; Phase 3 makes the read-through (not re-grepping) the step that passes sanitization - SKILL.md Step 5: sanitization is mandatory for public skills (was Optional); the semantic read-through is the method, the scan is a helper - SKILL.md Step 6: spell out what gitleaks does NOT cover (a green scan != sanitized) - Scrub the checklist's own examples, which ironically used real company / product / person / project names as "sanitization examples" -> generic placeholders Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(skill-creator): bump daymade-skill to 1.1.0 + add AI read-through as 5th sanitization layer - marketplace.json: daymade-skill 1.0.1 -> 1.1.0 (MANDATORY version bump for the skill-creator content change in this PR -- per CLAUDE.md "Updating Existing Skills") - CLAUDE.md: add layer 5 (AI semantic read-through) to the defense system. Layers 1-4 (CLAUDE rules / pre-commit / pre-push / gitleaks) are all keyword matching and structurally miss no-keyword private content; the read-through is the only gate that catches it. References the checklist SSOT rather than duplicating it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…the PDF Previews were created in the PDF's own directory. The self-check runs out-of-process (the caller Reads the PNGs), so the script cannot know when inspection is done and never cleans them up — the PNGs linger and pollute the working tree / git repo. Now written under $TMPDIR/pdf-creator-previews/<stem>/. Self-check is unchanged (the script prints the path to Read); the temp dir is reclaimed by the OS and never touches the repo. --no-preview behavior unchanged. - md_to_pdf.py: preview dir -> system temp dir (honors $TMPDIR) - SKILL.md: document new preview location - test_self_check_preview.py: regression guard (previews in temp, never next to PDF); isolate via $TMPDIR - marketplace.json: bump daymade-docs 1.0.1->1.0.2, marketplace 1.56.0->1.57.0 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…hub-and-spoke debugging-network-issues becomes the methodology hub: add a "Triage first" routing table that points known-domain symptoms (Tailscale/proxy fake-ip, Cloudflare, Windows AVD) to the dedicated skill. The 3 domain skills (tunnel-doctor, cloudflare-troubleshooting, windows-rdp) each gain a one-line "Methodology base -> debugging-network-issues" backlink. No domain content changed - pure cross-referencing, so a symptom like "git push: Connection closed by 198.18.x.x" routes straight to tunnel-doctor instead of re-deriving the methodology from scratch. Bumps: debugging-network-issues/cloudflare-troubleshooting/windows-rdp 1.0.0->1.0.1, tunnel-doctor 1.5.0->1.5.1, marketplace 1.57.0->1.58.0 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ardown Distilled from a real 14-agent due-diligence workflow. Takes a benchmark (founder/KOL/company/product) whose success looks inflated and produces a decision-oriented teardown: fan-out collection, adversarial L1-L4 verification (bubble-busting), attribution weighting (replicable vs luck/timing), then mapping the validated playbook onto the user's own resources. Edge over deep-research: adversarial debunking + attribution + self-mapping, not a neutral briefing. Inline orchestrator (spawns parallel agents). Reuses deep-research/osint/qcc for plumbing. Privacy: commissioner context only injected into the final mapping agent, never into external-search agents. - SKILL.md (100 lines) + 4 references (grading rubric / discipline traps / attribution-mapping / workflow template), all sanitized - marketplace.json: register skill v1.0.0, bump marketplace 1.58.0 -> 1.59.0 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ner P0 12 new StructuredDataREST methods (financials income/balance/cash-flow, TTM metrics & ratios, company profile, daily prices, dividends, revenue segments, entity-sentiment time series, co-mention graph) + BatchSearch (create/upload/poll/download, 50% off) + fields_values_to_records helper. Fix P0: company_screener filters must nest under "filters" (flat was silently ignored, returned an unfiltered universe); market_cap_lower_than field name. Docs: precise MCP-lossy thesis, SDK EOL 2026-12-31 note, 52x downgraded to a measured single point, expose official verify_ssl/proxy params, rc() rate-limit exclusion. All contract-tested against the official OpenAPI spec; gitleaks clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Doc-governance pass found two spots the prior commit's SKILL.md body missed (the references/ files were already correct): opening paragraph still said MCP 'exposes none of the /v1/* structured-financial endpoints' (now: MCP gives prose + tearsheets, not the machine-readable substrate); known-pitfalls daymade#6 still said screener filters are 'flat, not nested' (now: must nest under filters, flat is silently dropped -> unfiltered universe). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ted)
Doc-governance follow-up: live-ran analyst_ratings(NVDA) + price_target(NVDA) against the trial API (0 chunks) — both return {results}, so bumped verification level L3 (doc-confirmed) -> L4 (runtime-verified) in verified_api_signatures.md + the two docstrings. Also confirmed analyst_estimates quarter limit=30 really 400s (validates the existing <=20 note), verify_ssl passthrough constructs, and rc() excludes 429/rate-limit from retry.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ve-tested) End-to-end batch test caught a real bug: upload_input sent Content-Type application/octet-stream, but the S3 presigned URL signs content-type=application/jsonl (visible in the URL query) -> 403 SignatureDoesNotMatch. Fix: parse content-type from the presigned URL query and match it; add ProxyError retry (the proxy occasionally 503s on the S3 host). Live-tested 403->200; create/upload/poll(pending->processing) now L4. download_results stays unverified -- smart batch processing ran >10min without reaching completed; it is a standard S3 GET. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…, add date_range Live test (doc-governance follow-up) found connected_entities took an entity_categories param that does NOT exist on the co-mentions/entities endpoint (OpenAPI spec confirmed) — passing it was silently ignored, so 'category filtering' returned all 6 groups. Removed it. Replaced with date_range -> query.filters.timestamp (ANSI date-time, which the spec DOES support) — live-tested, returns the co-mention graph for the window. Results are already grouped by category; read the group you want. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…rget L3->L4 Doc-governance pass: the endpoint coverage table in rest_ext.py's module docstring still marked analyst_ratings/price_target as L3, contradicting the method docstrings + verified_api_signatures.md already bumped to L4 (live-tested 2026-05-31). Synced the table. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…point table Doc-governance root-cause fix: the module-docstring 'covered endpoints' table copied verified_api_signatures.md (endpoint list + verification levels) and had drifted twice this session — it listed only the original 7 endpoints (missing the 12 added) and separately needed the L3->L4 fix. Replaced the copied table with a one-line coverage overview pointing to verified_api_signatures.md as the single source for the full list / signatures / L3-L4 levels. Removes the copy that caused the drift. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Automated repository environment configuration, fault diagnosis, and repair for non-technical users. Includes SessionStart hook initialization, counter-review workflows, and git history sanitization. New files: - auto-repo-setup/SKILL.md - auto-repo-setup/scripts/check_env.py - auto-repo-setup/scripts/init_session_start_hook.py - auto-repo-setup/scripts/sanitize_history.sh - auto-repo-setup/references/git_safety.md - auto-repo-setup/references/pii_guard.md - auto-repo-setup/references/onboarding_template.md Updated: - .claude-plugin/marketplace.json (bump to v1.60.0) - CHANGELOG.md - CLAUDE.md (skill count 53→54) - README.md (skill count 52→53, add skill description) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Merged by Claude Code at user request.
Register bloom-tutor in marketplace.json plugins[] (source: ./bloom-tutor, category: productivity) and document it in README. A prompt-only skill (SKILL.md + 4 references) implementing a one-on-one Socratic tutor based on Benjamin Bloom's "2 Sigma Problem" (1984): generates one course document at a time, iterates on learner feedback toward +2 sigma mastery. Validated with skill-creator/quick_validate.py (Skill is valid!). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
daymade
force-pushed
the
main
branch
2 times, most recently
from
915a8c6 to
9c9615e
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Adds bloom-tutor, a prompt-only skill that turns Claude Code into a one-on-one Socratic tutor based on Benjamin Bloom's "2 Sigma Problem" (1984). Each subject is a self-contained folder; the tutor generates one course document at a time, waits for the learner's written feedback, and iterates the syllabus toward +2σ mastery. Documents (syllabus + numbered articles + an auto-generated summary) are the primary learning vehicle; conversation only confirms state. Responds in Chinese.
Source skill lives at https://github.com/Li-Evan/Bloom/tree/main/skills/bloom-tutor (138★, MIT).
Type of Change
Changes Made
bloom-tutor/(SKILL.md + 4 references: syllabus, articles, logging, summary).claude-plugin/marketplace.json(source: ./bloom-tutor,category: productivity,strict: false)Validation
marketplace.jsonparses as valid JSON (43 plugins total)skill-creator/scripts/quick_validate.py ./bloom-tutor→ "Skill is valid!"name+descriptionwith clear activation triggersNotes
Pure prompt-based skill — no scripts, no external API keys, more than a single SKILL.md (4 bundled references). Non-promotional, standalone value.