New and Recently Updated Skills
Verified Agent Skills sorted by source update date. This is where newly published SKILL.md files should surface first.
Dsh Code Review
Use when reviewing a pull request in the deepseek-harness repo — orients the reviewer to this codebase's standards (AGENTS.md conventions, defensive patterns, ADRs, quality gates) and the review-specific checks that code alone can't show
Dsh Doc Site Sync
Use when publishing, updating, moving, or removing DeepSeek Harness documentation website pages; editing website/docs.ts mappings or navigation; diagnosing a page missing from the VitePress site; fixing projected documentation links; or running the docs:dev, docs:check, and doc-sync workflow after website-content changes.
Dsh Doc Standards
Use when writing, moving, reviewing, or auditing documentation in the deepseek-harness repo — choosing hierarchy and detail, separating tutorials from references, checking tutorial progression, trimming doc slop, responding to a verify-doc-budgets failure, or requests like "improve the docs", "audit the docs", "where should this be documented", or "this doc is too long".
Dsh Find Simplifications
Use when working in the deepseek-harness repo to find non-obvious simplification candidates, write proposed Agent Notes or inline TODO/FIXME/XXX notes, audit or coalesce superseded Agent Notes, or fold worthwhile simplification ideas from another PR; especially for dead, duplicated, speculative, over-built, added-then-removed, or hand-rolled-where-a-dependency-exists surfaces.
Dsh Merging Stacked Prs
Use when landing a stack of dependent GitHub PRs (A ← B ← C, where each bases on the one below) onto master, merging a PR whose base is another open PR's branch, or whenever a request mentions "stacked PRs", "PR stack", "dependent PRs", or merging several related PRs in sequence. Requires every same-repository dependency chain to use GitHub's official stacked-PR feature before landing so GitHub owns stack-wide rules, CI, ordering, retargeting, and merge state.
Dsh Pre Push Checks
Use before pushing, force-pushing, marking ready for review, or claiming checks pass on a deepseek-harness branch, and immediately after gh stack sync publishes rewritten branches, to select the smallest tests and checks that cover the outgoing or just-published diff without reflexively running the full repository suite.
Dsh Prose Standard
Use when writing, reviewing, restoring, trimming, or auditing prose in the deepseek-harness repo, including deciding where documentation or comments are required across Markdown, JSDoc, code and test comments, prompts, descriptions, diagnostics, and CLI or UI strings.
Dsh Translate Docs
Manually run the extended DeepSeek Harness bilingual-document workflow, including generated briefings, delegated prose translation, whole-document translation, and scoped pairing verification.
Dsh Trim Cot Leakage
Use when auditing or fixing prose that reads like a leaked reasoning transcript — dead design-session citations such as (decision N), audit item codes, or §N of uncommitted drafts; change narration such as "used to", "no longer", "this cut"; stack or review vantage ("a later PR in this stack", "rejected in review"); reviewer-addressed justifications; control-flow narration; or hedged planning residue in comments, JSDoc, docs, or Agent Notes.
Editing Cordis Compositions
Use when creating, changing, or validating a Cordis composition for this harness — writing or editing an agent preset, adding or removing a plugin row, deciding whether something belongs to the host composition or to one session, checking whether a preset you authored actually mounts, or diagnosing a row that mounted but contributed nothing.
Record Browser Gif
Record browser or Web UI interaction demos as optimized GIFs using the available built-in browser, state-based frame capture, and deterministic encoding, then publish to a dedicated assets branch when the task includes attaching the GIF to a pull request. Use when asked to make, record, or generate a GIF that demonstrates a browser workflow, and for every pull request that changes product-user-visible GUI behavior, which MUST include a GIF recorded from the pull request's real server and model flow.
Career Ops
AI job search command center -- evaluate offers, generate CVs, scan portals, track applications. Use when the user pastes a job URL or JD, asks to scan portals, generate a CV/PDF, track applications, prepare for interviews, draft outreach/emails, or run any career-ops mode.
Career Ops Plugin Apify
How to scan a job source through an Apify actor as a keyed provider.
Career Ops Plugin Gmail
How to pull job leads from a Gmail label into the career-ops pipeline.
Career Ops Plugin {{NAME}}
How to use the {{NAME}} plugin and the data it produces.
Career Ops Plugin Notion
How to mirror the career-ops tracker to a Notion database and read records back as job leads.
Alef Generated Bindings
Alef-managed generated bindings in packages/* and binding crates — the regeneration workflow (task alef:generate / alef:format), Go DocumentExtractor→Extractor rename mapping, the git-diff freshness check, and the key alef.toml sections. Load before editing anything under packages/* or a binding crate, or when regenerating, formatting, or verifying Alef output.
Api Server Mcp
REST API server and MCP protocol integration
Batch Extraction
Use when extracting from many files at once with shared config, bounded parallelism, per-file overrides, and error recovery. Covers the `batch` command, `--file-configs`, `--max-concurrent`, and output layout.
Chunking
Use when splitting extracted text into chunks for LLM context windows or RAG ingestion. Covers chunk size, overlap, markdown/yaml/semantic chunkers, tokenizer-based sizing, and the standalone `chunk` command.
Chunking Embeddings
Chunking, embeddings, and RAG pipeline integration
Config Loading Precedence
How Xberg resolves configuration — CLI-mode and server/MCP-mode precedence orders, config file auto-discovery (xberg.{toml,yaml,json}), field-level inline JSON merge (merge_json_into_config), config file formats, and apply_extraction_overrides. Load when adding a config flag or env var, changing config precedence, or debugging why a setting is or isn't taking effect.
Crate Structure
The Xberg workspace layout — the version source of truth (root Cargo.toml [workspace.package] version), the crates/ workspace members (core lib, CLI, FFI, per-language binding crates, OCR crates), the out-of-workspace binding packages under packages/ (PyPI, RubyGems, Composer, Go, Maven, NuGet, Hex), and the tools/ directory. Load when navigating the repo, deciding where code belongs, or wiring a new crate or binding package.
Extracting Keywords
Use when extracting keywords (YAKE/RAKE) from documents — and, secondarily, when detecting document language or generating embeddings for RAG and search. Covers the keyword config (and its feature gating), `--detect-language`, and the standalone `embed` command with real flags.
Extracting Tables
Use when extracting tabular data from PDFs, spreadsheets, or images. Covers layout-aware table detection, table model selection, output formats (markdown / JSON cells), and known limits.
Extracting With Ocr
Use when extracting text from scanned PDFs, photographed pages, or images that have no embedded text layer. Covers OCR backends, language packs, force-OCR, and performance tuning.
Extraction Pipeline Patterns
Document extraction pipeline architecture and patterns
Feature Flag Policy
Cargo feature flags for crates/xberg — ORT-incompatible targets (WASM, Android x86_64 emulator), type-only and tract inference companion features, WASM/Android-safe variants, PDF backend, mutually-exclusive ORT variants, platform-conditional deps, aggregate feature sets, and build profiles. Load when adding, wiring, or debugging a Cargo feature, or when reasoning about what compiles on WASM/Android/Windows/macOS-intel targets.
Format Specific Extraction
Format-specific document extraction workflows
Mime Detection Routing
MIME type detection and extractor routing in core/mime.rs and core/extractor/bytes.rs — the extension→EXT_TO_MIME→validate→registry→extractor flow, key detection functions, the 115 case-insensitive extension map, priority-based registry selection, wildcard MIME families, and how to add a new MIME type. Load when adding a format, wiring an extractor to a MIME type, or debugging why a file routes to the wrong (or no) extractor.
Picking A Format
Use when choosing an output format for extracted documents — text, markdown, djot, html, or JSON. Maps consumer (LLM, parser, archive) to the right `--format` / `--content-format` pair.
Plugin Architecture Patterns
Plugin architecture, registration, and trait patterns
Release Versioning
How xberg versions are synced and released — Cargo.toml is the single source of truth, `task version:sync` propagates it to alef-managed binding manifests AND the integrations under integrations/, which are versioned and published in lockstep with core (including -rc.N). Load before bumping a version, editing the version-sync task, or touching an integration's version/xberg dependency.
Wasm Constraints
WASM build constraints for the crates/xberg-wasm crate — the wasm-target feature set, no-tokio sync-only internal APIs, the mandatory SyncExtractor trait for WASM-compatible extractors, the 2 MB HTML size limit, size-optimized build config (opt-level="z"), and the async-wrapper/sync-internal API pattern. Load when building for wasm32, adding or modifying a WASM-compatible extractor, or debugging WASM build/runtime failures.
Xberg
Extract text, tables, metadata, and images from 101 document formats (PDF, Office, images, HTML, email, archives, academic) using Xberg. Use when writing code that calls Xberg APIs in Python, Node.js/TypeScript, Rust, or CLI. Covers installation, extraction (sync/async), configuration (OCR, chunking, output format), batch processing, error handling, and plugins.
Budget Optimizer
Use when the user asks to "allocate my influencer budget", "optimize spend across tiers", or "compare budget scenarios"; produces a tier/platform/content allocation table, ROI and CPM/CPE projections, scenario comparisons, and mid-campaign reallocation moves. Not for building the full campaign plan — use campaign-planner. 达人预算分配/投放预算优化
Report Generator
Use when the user asks to "create a campaign report", "build an executive summary", or "deliver client results"; produces audience-tailored influencer marketing reports (executive, client, internal team) with data tables, narrative, key learnings, and recommendations. Not for raw metric computation — use performance-analyzer. 达人营销报告/结案汇报
Ad Account Auditor
Use when auditing a paid ad account for incremental contribution, wasted spend, or measurement integrity before scaling; runs a typed 20-item ROAS profile with verified vetoes and a SHIP/FIX/BLOCK/UNDECIDED gate on own exported data. Not for campaign structure design — use campaign-architect; not for creative production — use ad-creative-builder. 付费广告账户审计/ROAS评分
Ad Creative Builder
Use when the user asks to "write ad copy", "generate RSA headlines", or "build ad creative at volume"; produces ad units — RSA headlines/descriptions, hooks, and an angle matrix — message-matched to the destination landing page. Not for scoring an ad account — use ad-account-auditor; not for the post-click page — use landing-optimizer; not for organic articles — use content-writer. 广告创意/广告文案/RSA标题
Ad Test Designer
Use when the user asks to "design an A/B test", "set up a creative/landing test", "run an incrementality test", or "is this result statistically and practically material?"; produces a hypothesis, variant matrix, sample-size/duration/power plan, and a documented effect/uncertainty read from own exported results. It applies only a precommitted owner-approved action rule; the statistical helper never chooses a business action. Not for producing variants — use ad-creative-builder; not for reading back one shipped change — use paid-measurement-loop. 广告AB测试设计/实验设计/显著性判定/增效测试
Attribution Reconciler
Use when platform-reported conversions disagree with GA4/ecommerce, when you suspect Meta and Google are double-counting the same sales, or for a standing (monthly) reconciliation workbook that de-dups stacked credit against an order-ID truth set, normalizes attribution windows and currency, compares attribution models, and reads incrementality from a geo/holdout test. Not for the point-in-time R2 veto or RQS gate — use ad-account-auditor; not for the ROI/ROAS ratio math itself — use roi-calculator; not for organic dark-social share attribution or GA4 direct-traffic decomposition — use dark-social-attributor. 付费广告归因对账/去重/增量
Audience Mapper
Use when the user asks to "analyze my target audience", "build an audience profile for influencer targeting", "research a niche community", or "deep-dive a subculture before partnering with creators"; in audience mode produces demographic/psychographic profiles, a platform-priority matrix, named personas, and an influencer-selection criteria set, and in niche mode produces a community map, culture decode (language/norms/taboos), key-voice tiers, a Brand Fit Score, and a phased entry strategy. Not for finding specific creators to contract — use influencer-discovery; not for scoring a shortlist on Suitability — use fit-scorer. 目标受众画像/人群分析 · 细分社群/亚文化调研
Audience Segment Builder
Use when the user asks to "build audience segments from my customer list", "make value-based / lookalike seed lists", "set up exclusion / suppression segments", or "map audiences to funnel stages across platforms"; turns the user''s OWN customer/CRM/GA4 export into seed audiences, value-based lookalike SEED lists, exclusion/suppression segments, and a cross-platform funnel-stage targeting map, informing the ROAS A (Audience) dimension. Not for building account structure or match types — use campaign-architect; not for organic SERP intent — use keyword-research. 付费广告受众分群/种子人群/排除人群/相似人群种子
Bid Strategy Planner
Use when the user asks to "pick a bid strategy", "set a tCPA/tROAS target", or "plan the learning-phase entry"; produces a bid-strategy choice (tCPA / tROAS / max-conversions / manual CPC), the starting target math, a portfolio grouping map, and a learning-phase entry plan. Not for splitting the budget across campaigns — use budget-optimizer; not for in-flight pacing/scale moves — use budget-pacing-monitor; not for scoring the account — use ad-account-auditor. 出价策略/tCPA目标/tROAS/学习期
Brief Generator
Use when the user asks to "create an influencer brief" or "write a campaign brief"; produces a structured creator brief with deliverables, key messages, creative direction, timeline, disclosure rules, and compensation terms. Not for choosing how to split spend across creators — use budget-optimizer. 达人合作简报/创作者BF
Budget Pacing Monitor
Use when the user asks to "check pacing", "am I over/under-spending", "is this campaign on track to hit budget", or "why did spend spike/stall mid-flight"; returns a spend-vs-target-curve read, learning-phase status, an over/under-delivery call, and a reallocation trigger. Not for initial budget allocation — use budget-optimizer; not for choosing the bid strategy — use bid-strategy-planner; not for the RQS gate — use ad-account-auditor. 付费广告预算节奏监控/跑量过快过慢/在途配速
Campaign Architect
Use when the user asks to "plan my paid account structure", "pick Search vs PMax", "lay out ad groups / asset groups", or "audit paid-vs-organic cannibalization"; designs campaign-type selection, ad-group/asset-group layout, targeting + match types, negative/exclusion hygiene, and a paid↔organic overlap audit, and scores the ROAS A (Audience) dimension + structure. Not for computing the final RQS — use ad-account-auditor; not for budget split — use budget-optimizer; not for organic site architecture — use site-structure-optimizer. 付费广告账户结构/广告系列规划/否定关键词
Campaign Planner
Use when the user asks to "plan an influencer campaign", "build a campaign blueprint", or "launch a product with creators"; produces campaign objectives, platform and influencer-tier strategy, content requirements, a phased timeline, budget allocation, and KPI targets. Not for writing individual creator briefs — use brief-generator; not for the overall product-launch plan (tiering, calendar, press, community day) — use launch-tier-planner, which hands this skill the creator lane. A launch request that does not mention creators routes to the launch discipline, not here. 达人营销策划/种草方案