Skill_Seekers

[3.9.1] - 2026-08-02

Theme: Documentation and project-infrastructure release. No runtime code changed — the package is functionally identical to 3.9.0.

Added

• Sponsorship program (#442) — SPONSORSHIP.md (five tiers plus the rules), SPONSORS.md, and a sponsors.json single source of truth rendered into all 12 READMEs by scripts/render_sponsors.py. Adding a sponsor is one JSON edit instead of thirteen hand edits, and CI (--check) fails on drift. .github/FUNDING.yml gains GitHub Sponsors alongside Buy Me a Coffee.
• Sponsor URLs are policy-checked in code — the renderer refuses any sponsor link carrying utm_/ref/fbclid/gclid-style parameters, enforcing rule 4 of the published policy. This immediately caught 24 UTM-tagged links already present across the READMEs; all sponsor links are now clean.

Changed

• README restructured into a lean landing page — 1,371 → 424 lines, with the deep-dive material moved behind links to the docs/ pages that already covered it. Mirrored across all 11 translations (~16,500 → ~5,100 lines total), each keeping its machine-translation notice and language switcher.
• Repo root decluttered — BULLETPROOF_QUICKSTART.md → docs/getting-started/, the dated DOCUMENTATION_AUDIT_REPORT_2026-05-30.md → docs/archive/historical/, with every reference repointed.
• QWEN.md reduced from 719 to 64 lines — it duplicated AGENTS.md while advertising three different version numbers at once (v3.6.0, 3.3.0, and "17+ source types" beside "18 source types"). It is now a pointer to AGENTS.md that keeps the essential commands and conventions inline and carries no hardcoded version.

Fixed

• Stale and incorrect README claims corrected against the codebase — version badge 3.7.0 → 3.9.0; the documented scan --quick|--comprehensive|--enhance flags do not exist and are now create --preset quick|standard|comprehensive; export targets "16 formats"/"21 platforms" → 22; CLI reference "all 20 commands" → 19; the install-agent table 15 → 19 entries; workflow presets "24+" → 68; and the dead # skill-seekers list-configs line removed.
• Three overlapping troubleshooting guides unified — root TROUBLESHOOTING.md (485 lines), docs/TROUBLESHOOTING.md (1,102) and docs/user-guide/06-troubleshooting.md (108) shared only 2 of 22 headings, so none was a stale copy. docs/TROUBLESHOOTING.md is now the single comprehensive reference (1,398 lines) with the five sections that existed only at root merged in; the user-guide chapter stays as the short entry in the numbered series.
• 92 broken relative links repaired across the docs tree (109 → 17) — targets relinked to their real locations, case/separator mismatches fixed (integrations/cursor.md → CURSOR.md, advanced/api-reference.md → reference/API_REFERENCE.md), and links to never-written docs unlinked rather than left as 404s. The remaining 17 are intentional (template placeholders, illustrative generated-output samples, archived snapshots).
• Version stamps refreshed in 28 docs files (14 English + 14 zh-CN) from v3.6.0 to v3.9.0. Historical references in MIGRATION_GUIDE.md and "added in vX.Y.Z" notes are deliberately untouched.
• test_cli_paths.py no longer pins a version literal — it asserted "3.9.0" and broke the moment the dev version moved, the same fragility already fixed in test_package_structure.py. It now tracks skill_seekers.__version__.
• The published MCP Docker image was broken and is now fixed — Dockerfile.mcp installed the MCP SDK with a bare pip install mcp, bypassing the mcp>=1.25,<2 pin in pyproject.toml. Once mcp 2.0.0 shipped, that resolved to 2.x, where FastMCP moved out of mcp.server, so the container died on startup with cannot import name 'FastMCP'. It now installs via the [mcp] extra so the pin applies. Verified by building the image and confirming the server answers /health in ~4s on mcp 1.29.0.
• **Docker MCP image smoke test po

[3.9.0] - 2026-07-29

Added

• HTML parser fallback for broken markup (#96, task F1.4) — new parse_html() helper (cli/html_parsing.py) keeps html.parser as the primary parser (well-formed pages stay byte-identical) but retries with lxml (if installed) then html5lib when parsing raises or yields a tag-free tree for tag-bearing markup, so severely malformed pages no longer scrape as empty. Wired into all web-scraper parse sites (doc_scraper.py) and local HTML file parsing (html_scraper.py); html5lib added as a core dependency so the fallback always exists.
• Doc scraper retries transient network failures (#97) — scrape_page/scrape_page_async now wrap fetches in retry_with_backoff, so a connection blip or 5xx no longer silently drops the page and ships an incomplete skill. Only transient faults retry (connect/timeout errors, 5xx); 4xx is a definitive answer and returns un-retried. Attempts configurable via the new max_retries config option (default 3; 1 disables retrying).
• MCP fetch_config retries transient network failures (#92) — the registry list/detail/download GETs go through _get_with_retry (3 attempts, exponential backoff); 4xx like "config not found" still returns immediately.
• Whisper transcription fallback (Tier 2) implemented (#420) — transcribe_with_whisper() was a placeholder that always raised, so local videos without subtitle files produced no transcript despite the documented 3-tier fallback. Now implemented with faster-whisper (device auto, --whisper-model size, language hint, per-segment confidence), and get_transcript() logs the actual transcription error instead of swallowing it.
• MiniMax image OCR + registry-driven multimodal provider support (#423) — MiniMax joins the AI provider registry for text enhancement and vision OCR (MINIMAX_API_REGION global_en/cn_zh, MINIMAX_API_PROTOCOL openai/anthropic — China-issued keys no longer 401 against the global endpoint). API_PROVIDERS entries now declare wire protocol and supports_images, _call_api branches on protocol (not provider name), and AgentClient.call_with_image() serves all image-capable providers with SKILL_SEEKER_VISION_PROVIDER auto-detection for video_visual frame OCR.
• word (docx) source in the scrape_generic MCP tool (#41) — the converter existed in CONVERTER_REGISTRY but the MCP surface never listed it, so MCP clients had no way to scrape .docx sources.

Changed

• Token-lean GitHub issue defaults (#169) — GitHub-sourced skills no longer bundle a repo's full closed-issue history by default: issue_state defaults to "open" (was "all"), max_issues to 20 (was 100), and labels/milestones are emitted only with the new include_issue_labels/include_issue_milestones flags (default off). issues.md omits empty sections and no longer prints "No labels" stubs. All defaults overridable.

Fixed

• Unified multi-source skip_scrape now rebuilds from cache — UnifiedScraper reloads each configured source from .skillseeker-cache/cached extraction files before conflict detection and build, so cached unified configs no longer re-scrape the network (#405).
• Quality completeness checks no longer match inside code blocks (#229) — SkillQualityChecker._check_skill_completeness ran its workflow-step/prerequisites/error-handling regexes against raw SKILL.md, so a # Step 1: comment in a fenced example counted as workflow guidance. Fenced blocks and inline code are stripped before matching; prose detection unchanged.
• Pattern-detector precision: getters are not factories, Java Singletons are detectable (#425) — FactoryDetector matched creation keywords as substrings, so a plain Java POJO scored Factory from its getters and a canonical getInstance() Singleton was misclassified as Factory. Creation verbs now match as word-boundary prefixes with accessor prefixes (get/set/is/has) excluded, and SingletonDetector recognizes c

[3.8.0] - 2026-06-15

Theme: The Grand Unification refactor (one build pipeline, one AI transport, one parser definition, in-process MCP tools — see docs/UNIFICATION_PLAN.md) with a 13-bug audit fix-up and a real-world end-to-end CLI testing pass — plus MiniMax-M3, model selection, registry-driven platform targets, a Windows subprocess fix, and codebase de-duplication.

Fixed

• Codebase skills built from a config no longer drop the API reference + dependency graph — the local source carried api_reference and dependency_graph, but the unified builder never wrote them out, so a skill built from a scan-emitted *-codebase.json shipped without its API reference (it was stranded in the scrape cache). Both are now promoted into references/codebase_analysis/<source>/ and linked from the index.
• create ./path defaults to deep analysis — local codebase create defaulted to surface depth, producing an empty code_analysis.json and a misleading "analyzed 0 files" log; it now defaults to deep (matching the scraper default and the scan-config path), with an explicit --depth still taking precedence.
• package works non-interactively — the quality-gate prompt raised EOFError when stdin wasn't a TTY (CI/pipes); it now auto-proceeds on a non-TTY or with the new --yes/-y flag, while an interactive terminal still prompts.
• quality prints the score — the standalone command saved quality_report.json but printed nothing; it now shows the score/grade summary, and the report serializes metric levels as "info"/"warning" instead of the Python enum repr "MetricLevel.INFO".
• doctor no longer miscounts GITHUB_TOKEN as an AI provider key — the API-keys check now names which keys are set, so a bare GitHub token isn't misread as a provider key being configured.
• estimate <url> gives an actionable error — passing a URL (which create accepts) printed a bare "Config file not found"; it now explains that estimate takes a config file and points at create --dry-run, and exits non-zero.
• Web sitemap discovery fails fast on unreachable hosts — the pre-crawl sitemap probes used a single scalar timeout; they now use a (connect, read) timeout so an unreachable host doesn't block the full window before the crawl starts.
• Fork-bomb guard covers the primary LOCAL enhance path — _run_agent_command now marks SKILL_SEEKER_ENHANCE_ACTIVE in the spawned agent's environment (including terminal-mode scripts) and run() refuses nested spawns, so a local agent enhancing a skill can no longer recursively launch enhancement.
• --dry-run and --output honored for unified configs — create skipped injecting both into UnifiedScraper (every other source type got them); dry-run now previews and returns without creating directories, for legacy and unified configs alike.
• Trailing-slash --output no longer leaks intermediates into the packaged skill — SkillConverter resolves skill_dir once and strips trailing separators, so --output out/x/ can't place _extracted.json inside the skill directory.
• snake_case config classification — config_extractor._path_has_word used \b, which never matches inside snake_case names (app_db.yaml); explicit lookarounds fix detection while keeping dbeaver/blog false positives excluded.
• How-to guides no longer come out empty for control-flow-wrapped tests — step extraction recursively descends with/for/if/try blocks in source order, and empty AST results fall back to the heuristic instead of emitting empty guides.
• Slack 429 retry exhaustion logs a truncation warning (matching the Discord path) instead of silently breaking pagination.
• Kimi CLI output parsing no longer swallows unknown records — record boundary is any CamelCase constructor, so new record types (ToolCallPart, …) can't leak internals into extracted text.
• scan re-runs no longer churn .archived/ — canonical-named files fetched i

[3.7.0] - 2026-05-30

Theme: AI-driven project knowledge base (skill-seekers scan) — bootstrap a complete skill set for a project in one command, with safety/observability/coverage hardening throughout.

Added

• skill-seekers scan <dir> command (#327) — point at any project; an AI agent inspects manifests, README, Dockerfile/CI, sampled source files (first 2 KB each), and the git remote, then emits one Skill Seekers config per detected framework plus a <project>-codebase.json for the project's own code. Each config stamped with metadata.detected_version so re-scans report added / version-bumped / removed dependencies. Internationalized canonical-name resolver (CJK + EU language suffixes) so detections like "Godot 引擎" resolve godot. Out-dir cache means re-scans reuse prior emissions and respect manual edits. Doctor-style report with pluralized counts and resolved / AI-generated / unresolved / archived breakdown.
• Coverage: scan recognizes ~50 manifest types (Pipfile, environment.yml, deno.json, flake.nix, Chart.yaml, stack.yaml, deps.edn, dune-project, BUILD.bazel, …) and walks src/lib/app/cmd/crates/packages/apps/services/backend/frontend plus root-level files (catches Django, flat-layout Python, Go, Rust workspaces, JS monorepos).
• Cost + safety flags: --max-ai-generations N (default 10) caps unbounded AI generation for monorepos; --dry-run previews what would be emitted without writing or invoking AI; --probe-urls HEAD-probes AI-generated URLs with retry-on-404; --no-fetch / --no-generate / --no-publish-prompt for offline / CI use.
• Community submission (opt-in): freshly AI-generated configs can be submitted to the community registry via a native-async flow. Pre-checks GITHUB_TOKEN, idempotency-guards against duplicate issues, retries transient failures with backoff.
• Archival: configs that disappear from detections are moved (not deleted) to out_dir/.archived/<UTC-timestamp>/ so the user never loses hand-edited work and out_dir stays clean.
• Docs: new docs/getting-started/05-scan-a-project.md; entries in README, FAQ, CLI Reference, Feature Matrix, Config Format, Environment Variables, and the Quick Start cross-link.

Changed

• CLI dispatch unified (#327) — scan and doctor now consume the parsed-args namespace directly via Command(args).execute() instead of building a second argparse.ArgumentParser. Eliminates the _reconstruct_argv hack for these commands; remaining ~14 commands flagged for migration.
• Config schema: detected_version lives under metadata.detected_version (alongside metadata.version for the config-schema version) rather than at top level. Backwards-compatible reader; old top-level placements migrate on next stamp.
• SourceDetector.CODE_PROJECT_MARKERS is now public (was _CODE_PROJECT_MARKERS); cross-module callers no longer reach into a private attribute.

Fixed

• Correctness (#327) — diff layer keyed by stable filename slug instead of internal config name (eliminates phantom add/remove churn); resolve_config_path lookups now append .json so local-disk + user-dir paths actually find files; out-dir cache prevents redundant API/AI calls on re-scan; lowercase filename slugs prevent duplicate-file accumulation across runs.
• Safety (#327) — atomic JSON writes via os.replace so SIGINT mid-write can't corrupt a config and silently flip it to "removed" on the next scan; _safe_size guards stat() so a broken symlink in src/ no longer crashes the scan; AgentClient.call exceptions caught and logged; AI-generated config names rejected if they fail the registry regex; URL probe catches AI hallucinations of base_url before writing.
• Observability (#327) — logging.basicConfig in scan so logger.warning/error reaches the user (was silently dropped); non-zero exit code when no configs and no codebase config were emitted, so