Results for “l-eval”

Turn a refined research proposal or method idea into a detailed, claim-driven experiment roadmap. Use after `research-refine`, or when the user asks for a detailed experiment plan, ablation matrix, evaluation protocol, run order, compute budget, or paper-ready validation that supports the core problem, novelty, simplicity, and any LLM / VLM / Diffusion / RL-based contribution.

Remove refusal behaviors from open-weight LLMs using OBLITERATUS — mechanistic interpretability techniques (diff-in-means, SVD, whitened SVD, LEACE, SAE decomposition, etc.) to excise guardrails while preserving reasoning. 9 CLI methods, 28 analysis modules, 116 model presets across 5 compute tiers, tournament evaluation, and telemetry-driven recommendations. Use when a user wants to uncensor, abliterate, or remove refusal from an LLM.

Revenue leadership for B2B SaaS companies. Revenue forecasting, sales model design, pricing strategy, net revenue retention, and sales team scaling. Use when designing the revenue engine, setting quotas, modeling NRR, evaluating pricing, building board forecasts, or when user mentions CRO, chief revenue officer, revenue strategy, sales model, ARR growth, NRR, expansion revenue, churn, pricing strategy, or sales capacity.

Growth strategy for product-led and loop-driven growth systems. Use when building growth strategy, designing growth loops, planning acquisition channels, evaluating network effects, deciding when to scale, or coordinating product, engineering, data, and marketing around growth. For SEO audits use seo-and-aeo-strategy; for page/form CRO use conversion-rate-optimization; for A/B test design use ab-test-setup.

Fine-tunes and evaluates OpenVLA-OFT and OpenVLA-OFT+ policies for robot action generation with continuous action heads, LoRA adaptation, and FiLM conditioning on LIBERO simulation and ALOHA real-world setups. Use when reproducing OpenVLA-OFT paper results, training custom VLA action heads (L1 or diffusion), deploying server-client inference for ALOHA, or debugging normalization, LoRA merge, and cross-GPU issues.

Fast in-memory DataFrame analytics with Polars — lazy evaluation, parallel execution, and an Apache Arrow backend for datasets that fit in RAM. Use when pandas is too slow but data still fits in memory, for 1-100GB datasets, ETL pipelines, or a faster pandas replacement. For larger-than-RAM data prefer dask or vaex. Part of the AlterLab Academic Skills suite.

Multi-agent collaboration plugin that spawns N parallel subagents competing on the same task via git worktree isolation. Agents work independently, results are evaluated by metric or LLM judge, and the best branch is merged. Use when: user wants multiple approaches tried in parallel — code optimization, content variation, research exploration, or any task that benefits from parallel competition. Requires: a git repo.

Multi-agent collaboration plugin that spawns N parallel subagents competing on the same task via git worktree isolation. Agents work independently, results are evaluated by metric or LLM judge, and the best branch is merged. Use when: user wants multiple approaches tried in parallel — code optimization, content variation, research exploration, or any task that benefits from parallel competition. Requires: a git repo.

Multi-agent collaboration plugin that spawns N parallel subagents competing on the same task via git worktree isolation. Agents work independently, results are evaluated by metric or LLM judge, and the best branch is merged. Use when: user wants multiple approaches tried in parallel — code optimization, content variation, research exploration, or any task that benefits from parallel competition. Requires: a git repo.

Multi-agent collaboration plugin that spawns N parallel subagents competing on the same task via git worktree isolation. Agents work independently, results are evaluated by metric or LLM judge, and the best branch is merged. Use when: user wants multiple approaches tried in parallel — code optimization, content variation, research exploration, or any task that benefits from parallel competition. Requires: a git repo.

Multi-agent collaboration plugin that spawns N parallel subagents competing on the same task via git worktree isolation. Agents work independently, results are evaluated by metric or LLM judge, and the best branch is merged. Use when: user wants multiple approaches tried in parallel — code optimization, content variation, research exploration, or any task that benefits from parallel competition. Requires: a git repo.

Use this skill when the user requests to review, analyze, critique, or summarize academic papers, research articles, preprints, or scientific publications. Supports comprehensive structured reviews covering methodology assessment, contribution evaluation, literature positioning, and constructive feedback generation. Trigger on queries involving paper URLs, uploaded PDFs, arXiv links, or requests like "review this paper", "analyze this research", "summarize this study", or "write a peer review".

Generate academic referee reports for economics/finance papers, followed by a 150-word letter to the editor with recommendation (Reject / Major Revision) and Kai Wu signature. Two modes (normal / high-level), configurable number of comments; recommendation choice drives evaluation tone (Reject → negative, Major Revision → neutral). Trigger when user says "referee report" / "write referee report" / "审稿报告" / "写审稿意见" / "generate referee report" / "review this paper".

Multi-agent collaboration plugin that spawns N parallel subagents competing on the same task via git worktree isolation. Agents work independently, results are evaluated by metric or LLM judge, and the best branch is merged. Use when: user wants multiple approaches tried in parallel — code optimization, content variation, research exploration, or any task that benefits from parallel competition. Requires: a git repo.

Give your agent a budget, a target, and a deadline — it does the rest. Orchestrates DSL + Opportunity Scanner + Emerging Movers into a full autonomous trading loop on Hyperliquid. Race condition prevention, conviction collapse cuts, cross-margin buffer math, speed filter. 3 risk profiles: conservative, moderate, aggressive. Use when setting up autonomous trading, creating a trading strategy, or running a scan-evaluate-trade-protect loop.

Generate 5–10 business idea candidates from a blank page or a founder's domain context — using pain mining, jobs-to-be-done, trend × capability mapping, constraint relaxation, adjacency search, and founder-market-fit prompts. Each candidate is a structured idea card (segment, JTBD, current alternative, why-now, distribution wedge, monetisation, "feels like"). Load when the user asks to generate business ideas, brainstorm startup ideas, find ideas to work on, says "what business should I start", "give me startup ideas", "I don't know what to build", "ideate ventures", "blank-page idea generation", "find me a startup idea", "explore business opportunities". Sub-skill of `venture-exploration`. Hard-bans "Uber for X" / "AI for X" with no specific JTBD, "everyone" segments, and idea cards missing any of the 7 required fields. Does NOT design or evaluate ideas generated — for that use `idea-evaluation`.

4-stage funnel that screens all 500+ Hyperliquid perps down to the top trading opportunities. Scores setups 0-400 across smart money, market structure, technicals, and funding. BTC macro filter, hourly trend gate (counter-trend = hard skip), cross-scan momentum tracking. Near-zero LLM tokens — all computation in Python. Use when scanning for new trading opportunities on Hyperliquid, evaluating setups, or checking market conditions.

Use to push call recordings into Rulebase over the REST API when the phone system has no native connection, in reviewed batches with a dry-run plan, resume and audit log. Trigger for "upload calls to Rulebase", "import call recordings", "backfill our call history into Rulebase", "get XCally calls into Rulebase", "Rulebase isn't seeing our calls", or bulk-ingesting audio for QA evaluation.

Determines whether to perform adrenal venous sampling (AVS) to lateralize aldosterone excess in patients with confirmed primary aldosteronism (PA) who are being evaluated for surgical treatment. Indicated when surgery is feasible and desired by the patient, or when subtype workup is planned for a surgical candidate, especially in those younger than 35 years with spontaneous hypokalemia, marked aldosterone excess, and unilateral adrenal lesions on CT.

Shipley-methodology federal proposal outline and section drafter. USE WHEN the user asks to draft a proposal volume, build an outline from the proposal_instruction ↔ evaluation_factor traceability (UCF Section L/M or equivalent for non-UCF — FAR 16 task orders, FOPRs, BPA calls, OTAs, agency-specific formats), generate a compliance matrix, write win themes, draft an executive summary, propose FAB (Feature → Advantage → Benefit) chains, identify discriminators, or 'respond to this RFP'. Pulls requirements, evaluation factors, instructions, customer priorities, and pain points from the active Theseus workspace KG and produces an evidence-cited draft. Also ships govcon HTML render templates under assets/ — hand the rendered content off to the `huashu-design` skill for PPTX / PDF / animation export. Format-agnostic — never assumes UCF section labels are present. DO NOT USE FOR clause compliance auditing only (use compliance-auditor) or extracting new entities (use govcon-ontology + the Theseus pipeline).

Use to verify that the chain from conversation to evaluation to decision is complete and reconstructible, so a QA score or a customer outcome can be explained months later. Trigger for "can we prove why this decision was made", "is our audit trail complete", "reconstruct how this score was produced", evidencing an AI-assisted decision, or an auditor asking how a conclusion was reached.

Use this skill for processing and analyzing large tabular datasets (billions of rows) that exceed available RAM. Vaex excels at out-of-core DataFrame operations, lazy evaluation, fast aggregations, efficient visualization of big data, and machine learning on large datasets. Apply when users need to work with large CSV/HDF5/Arrow/Parquet files, perform fast statistics on massive datasets, create visualizations of big data, or build ML pipelines that do not fit in memory.

Use this skill for processing and analyzing large tabular datasets (billions of rows) that exceed available RAM. Vaex excels at out-of-core DataFrame operations, lazy evaluation, fast aggregations, efficient visualization of big data, and machine learning on large datasets. Apply when users need to work with large CSV/HDF5/Arrow/Parquet files, perform fast statistics on massive datasets, create visualizations of big data, or build ML pipelines that don't fit in memory.

Reverse-engineers a federal RFP we received — given the SOW/PWS and evaluation criteria already in the Theseus KG, reconstructs the CO's hidden decision tree (upstream `sow-pws-builder` 6 scope blocks + 3 intake answers), surfaces hot buttons, ghost language, discriminator hooks, missing-section signals, and CPFF-form / Section-5 / QASP / Key-Personnel traps. USE WHEN the user asks "what scope decisions did the CO already make?", "reverse engineer this RFP", "what hot buttons are hiding in this PWS?", "where are the discriminator hooks?", "did they pick CPFF completion or term form?", "anything suspiciously missing?", or any variant of decoding CO intent. Pulls `requirement`, `deliverable`, `proposal_instruction`, `evaluation_factor`, `clause`, `performance_standard` from the active workspace KG and emits a JSON envelope feeding `proposal-generator`. DO NOT USE FOR proposal prose (`proposal-generator`), pricing (`price-to-win`), clause audit (`compliance-auditor`), or sub SOW (`subcontractor-sow-builder`).

Systematic methodology for auditing the Hermes Skills Hub (and external sources) to identify gaps, overlaps, redundancies, and conflicts when building a skill library for any project. Covers browsing, targeted searching, cross-referencing with web/GitHub, comparing candidates against existing inventory, verifying install safety, and planning installation order. Use when starting a new project and assembling tools, or when evaluating whether to add external skills beyond what's currently installed.

Use this skill for processing and analyzing large tabular datasets (billions of rows) that exceed available RAM. Vaex excels at out-of-core DataFrame operations, lazy evaluation, fast aggregations, efficient visualization of big data, and machine learning on large datasets. Apply when users need to work with large CSV/HDF5/Arrow/Parquet files, perform fast statistics on massive datasets, create visualizations of big data, or build ML pipelines that don't fit in memory.

Generate and critically evaluate grounded ideas about a topic. Use when asking what to improve, requesting idea generation, exploring surprising directions, or wanting the AI to proactively suggest strong options before brainstorming one in depth. Triggers on phrases like 'what should I improve', 'give me ideas', 'ideate on X', 'surprise me', 'what would you change', or any request for AI-generated suggestions rather than refining the user's own idea.

Use this skill for processing and analyzing large tabular datasets (billions of rows) that exceed available RAM. Vaex excels at out-of-core DataFrame operations, lazy evaluation, fast aggregations, efficient visualization of big data, and machine learning on large datasets. Apply when users need to work with large CSV/HDF5/Arrow/Parquet files, perform fast statistics on massive datasets, create visualizations of big data, or build ML pipelines that don't fit in memory.

Use to push back-office work into Rulebase over the REST API so it can be evaluated like a conversation, with a dry-run plan, idempotent external ids and an audit log. Trigger for "send our back-office work to Rulebase", "push work items", "QA our operations team", "get case work into Rulebase", disputes or KYC queues with no customer on the line, or duplicate work items appearing after a re-run.

Performs a gap analysis between two artifacts (a current state and a desired state) and produces a plain-language, stakeholder-readable report indexed by stable gap IDs. Use when the user wants to compare, evaluate, audit, or reconcile one artifact against another. Does not investigate runtime bugs — use investigate. Does not assess module-level architecture — use architectural-analysis. Does not research open-ended options with no second artifact to compare against — use research.

Select, pair, and implement typefaces for web projects. Use when the user mentions "font pairing", "which typeface", "line height", "responsive typography", "web font loading", "type hierarchy", "variable fonts", "FOUT/FOIT", or "typographic scale". Also trigger when choosing between system fonts and web fonts, optimizing font loading performance, or designing readable long-form content layouts. Covers readability evaluation, CSS implementation, and performance optimization. For overall UI design systems, see refactoring-ui. For dramatic typographic experiences, see top-design.

Out-of-core tabular analytics with Vaex for billion-row datasets that exceed RAM — lazy evaluation, fast aggregations, big-data visualization, and ML on a single machine. Use when working with large CSV/HDF5/Arrow/Parquet files, computing fast statistics on massive datasets, visualizing big data, or building ML pipelines that do not fit in memory. For distributed clusters prefer dask; for in-memory speed prefer polars. Part of the AlterLab Academic Skills suite.

Failure diagnosis, DAG mutation, and escalation engine for the WinDAGs meta-DAG. Receives failure information and quality vectors from the Evaluator. Classifies failures on four dimensions. Follows a five-level escalation ladder. Applies seven mutation types with saga compensation. Enforces BC-EXEC-002, BC-EXEC-003, BC-FAIL-002, BC-FAIL-005. Activate when operating as the Mutator role in the meta-DAG, when diagnosing node failures, when restructuring a DAG at runtime, or when deciding escalation level.

Apply meta-principles of software craftsmanship: DRY, orthogonality, tracer bullets, and design by contract. Use when the user mentions "best practices", "pragmatic approach", "broken windows", "tracer bullet", "software craftsmanship", "technical debt prevention", "prototype vs tracer bullet", or "code ownership". Also trigger when evaluating build-vs-buy decisions, designing estimation approaches, or choosing between reversible and irreversible architectural decisions. Covers estimation, domain languages, and reversibility. For code-level quality, see clean-code. For refactoring techniques, see refactoring-patterns.

Debug LLM applications using the Phoenix CLI. Fetch traces, analyze errors, structure trace review with open coding and axial coding, inspect datasets, review experiments, query annotation configs, and use the GraphQL API. Use whenever the user is analyzing traces or spans, investigating LLM/agent failures, deciding what to do after instrumenting an app, building failure taxonomies, choosing what evals to write, or asking "what's going wrong", "what kinds of mistakes", or "where do I focus" — even without naming a technique.

Integrate antennas into RF systems using MATLAB Antenna Toolbox and RF Toolbox. Covers impedance matching network design (L/Pi/Tee topologies, evaluation parameters, Richards transformation), measured antenna creation (E-field, directivity-only, EmbeddedE, ffsReader import), RF propagation and site planning (txsite/rxsite, coverage, SINR, ray tracing, link budget), and SAR estimation (birdcage+Phantom, conformalArray+Custom3D, direct EHfields). Use when the user wants to match an antenna, create a measuredAntenna, compute coverage or signal strength, perform ray tracing, or estimate SAR.