Sanoy24

Reproduce the τ²-Bench retail baseline, wrap the harness for traced scoring against dev and sealed held-out slices, and design 30+ Tenacious-specific adversarial probes with business-cost derivation. Use this skill whenever the user mentions τ²-Bench,…

Fetch public job posts, team pages, GitHub org activity, and press/funding signals for a Tenacious prospect via Playwright (or TinyFish), respecting robots.txt, the no-login rule, and the 200-company live-crawl cap for the challenge week. Use this skill…

Apply pragmatic Python and TypeScript/JavaScript code-quality conventions to the Tenacious Conversion Engine codebase — ruff + mypy + pytest for Python, eslint + prettier + tsc for TS/JS, with drop-in configs bundled. Use this skill whenever the user writes a…

Draft the two-page Tenacious decision memo (memo.pdf) addressed to the Tenacious CEO and CFO, with every numeric claim traceable to a trace file or published source via evidence_graph.json. Use this skill whenever the user mentions Act V, the final memo, the…

Draft outbound, reply, and nurture emails in the Tenacious Consulting voice that ground every claim in the hiring signal brief and respect bench/pricing guardrails. Use this skill whenever the user asks to write a cold email, warm reply, re-engagement…

Classify a Tenacious prospect into one of four ICP segments (recently-funded / mid-market-restructuring / leadership-transition / capability-gap) with a confidence score and an abstention option. Use this skill whenever the user needs to assign a prospect to…

Build the population-level market-space map over the full Crunchbase ODM sample — sector × size × AI-readiness cells, with hand-labeled precision/recall validation and top-cell ranking for outbound allocation. Use this skill whenever the user mentions the…

Generate a hiring signal brief and competitor gap brief for a Tenacious Consulting prospect from public data. Use this skill whenever the user mentions building outbound for Tenacious, enriching a prospect, the "hiring signal brief," "competitor gap brief,"…

Structure ablations, statistical comparisons, and evidence tracing for Tenacious-Bench results. Use when an agent is preparing Delta A, Delta B, or Delta C comparisons, held-out scoring traces, confidence intervals, cost or latency comparisons, evidence…

Build or extend Tenacious-Bench and its supporting Week 11 artifacts for the Sales Agent Evaluation Bench challenge. Use when an agent needs to turn Week 10 sales-agent traces, probes, style guidance, or public-signal inputs into a machine-verifiable…

Protect Tenacious-Bench split integrity and contamination resistance. Use when an agent is partitioning tasks, sealing held-out data, checking overlap between train and held_out, validating public-signal time windows, documenting contamination controls, or…

Create, expand, or review Tenacious-Bench tasks across the required authoring modes. Use when an agent is turning Week 10 traces, probe seeds, public-signal inputs, or sales artifacts into benchmark tasks, partition-ready records, metadata-rich JSON or JSONL…

Package Tenacious-Bench outputs into public, reviewer-friendly artifacts. Use when an agent is preparing a Hugging Face dataset or model card, datasheet, README, technical blog draft, publication checklist, executive memo support files, or…

Write, review, or refactor Python code for the Tenacious Bench project using clean, modern, and maintainable patterns. Use when an agent is adding or editing Python modules, CLI scripts, evaluators, dataset builders, contamination checks, training-data…

Design, tighten, or review machine-verifiable scoring rubrics for Tenacious-Bench tasks. Use when an agent is creating or revising schema fields, scoring evaluator logic, judge prompts, rubric dimensions, banned-phrase checks, grounding requirements, or…

Convert Tenacious-Bench artifacts into high-quality training inputs for Path A, Path B, or Path C. Use when an agent is formatting chat pairs, preference pairs, step-level labels, quality filters, or path-specific training partitions for LoRA, judge training,…

The security contract for this server — SSRF, DNS rebinding, redirect, and encoding defenses with the full threat matrix and checklist. Use whenever touching URL handling, DNS resolution, the HTTP client, redirects, or the security/ module.

How to test and validate this MCP server — unit tests for every SSRF threat, MCP Inspector checks, and external scanner validation. Use when writing tests, before shipping, or when verifying security correctness.

Best practices for writing MCP server tools in TypeScript (current SDK). Use when adding or editing a tool, its schema, description, annotations, error handling, or response format in this repo.

Reviews backend codebases for performance bottlenecks using an evidence-first, workload-driven methodology. Use when investigating latency, throughput, slow endpoints, database or query performance, N+1 queries, connection pool exhaustion, event-loop…