AI & ML

learningmatter-mit Bundle

Fine-tune Fairchem machine learning interatomic potentials (UMA, ESEN) on custom datasets.

learningmatter-mit Bundle

Identify and rank ligandable pockets on a protein structure or model using geometry (fpocket) or an ML predictor (P2Rank). Returns ranked pockets with lining residues, geometric center, volume, and a druggability score per pocket. Excludes docking; pair with drug-binding-site-definition or drug-docking-vina downstream. Use whenever the user has a protein but no binding-site information, asks about cryptic / allosteric / orphan pockets, needs to assess druggability, or wants to choose where to dock.

stas2-git Bundle

Use when creating evaluation cases, datasets, rubrics, golden prompts, red-team prompts, regression tests, LLM-as-judge criteria, or pass/fail checks for an agent workflow, skill, tool-using behavior, safety boundary, or production agent quality loop.

stas2-git Bundle

Use when deciding whether an agent capability should be a skill, MCP tool/server, A2A agent, A2UI interface, AP2/UCP commerce flow, ordinary script, or project instruction.

stas2-git Bundle

Use when working with untrusted external content such as web pages, emails, documents, PDFs, issues, tickets, tool outputs, retrieved context, or screenshots that may contain hostile instructions, prompt injection, hidden directives, data-exfiltration attempts, or policy-bypass requests.

stas2-git Bundle

Use when designing an agent architecture, choosing between tools/MCP/skills/multi-agent patterns, writing a production spec, planning evaluation, or turning a vague agent idea into a buildable system design.

stas2-git Bundle

Use when a task has too much source material, long documents, many references, noisy context, overlapping instructions, or repeated context rot; plans what belongs in static instructions, skills, references, scripts, retrieval, summaries, or not in context.

stas2-git Bundle

Use when an agent performs poorly, loops, calls the wrong tool, misses instructions, overuses context, hallucinates dependencies, or seems like it needs a better harness rather than a better model.

stas2-git Bundle

Use when designing human-in-the-loop approval flows for agent actions, including pending approvals, review payloads, allow/deny decisions, escalation rules, audit records, session resume, and what can be auto-approved versus blocked.

stas2-git Bundle

Use before giving an agent tools, MCP servers, file access, credentials, browser/session access, shell commands, write permissions, deployment authority, or any external action capability; plans scopes, read/write limits, approval gates, logging, and rollback.

stas2-git Bundle

Use when converting vague intent into production-grade agent/coding work with specs, BDD scenarios, failing tests first, small diffs, evidence-based debugging, review summaries, policy gates, and human approval boundaries.

stas2-git Bundle

Use when reviewing or debugging an agent run, trace, tool-call log, transcript, cost report, approval record, or suspicious final answer to inspect trajectory quality, context loaded, tool choices, retries, failures, cost, safety gates, and whether the final output is supported by the run.

stas2-git Bundle

Use after an agent mistake, unsafe action, bad answer, tool misuse, missed instruction, loop, user correction, production incident, or near miss to identify root cause, failed boundary, missing eval, missing guardrail, and the regression case that should prevent recurrence.

stas2-git Bundle

Perform a systematic STRIDE security assessment of the current codebase and architecture. Use when starting an implementation phase, reviewing agent tools or workflows, changing trust boundaries, or producing or updating threat_model.md.

rlawoals0529 Skill

Build a voice guide from someone's real writing, then draft in it without posting. Covers finding the tells that actually separate their writing from yours, evidencing every rule so the guide cannot drift into invention, keeping registers separate per surface, and the draft-and-hand-over rule. Use when an agent will write prose that goes out under a person's name, when a draft keeps coming back rewritten, or when setting up a new collaborator's voice. Triggers on "write as me", "make this sound like me", "in my voice", "draft the PR description", "this does not sound like me", "capture my writing style".

learningmatter-mit Bundle

Generate inorganic material structures using MatterGen, a diffusion-based generative model.

learningmatter-mit Bundle

Train a model to predict custom properties of molecules or periodic materials using pretrained SelfConditionedDenoisingAtoms (SCD) foundation models.

learningmatter-mit Bundle

Hierarchically decompose high-level scientific workflows (from literature or user-proposed) into executable sequences of existing SKILLs and MCP tools for the research plan.

ken3pei Bundle

Use when delegating work to another Herdr pane and the delegated agent should start inside a Treehouse-managed Git worktree. Default to pane; use a tab only when the user explicitly asks for tab-based delegation.

ken3pei Bundle

Review prompt token efficiency from local Claude Code and Codex prompt history. Use when the user invokes "/prompt-token-reviewer", asks to review prompt history, or wants to improve prompting efficiency.

alpayc Skill

Writes the operational document for a project — how to run it locally, its preconditions, how to reach a given state, how to debug it, the failure modes newcomers hit first — every command verified by running it. Use it when setup takes too long, or an agent must start an app it never has.

rlawoals0529 Skill

Orchestrate a fleet of coding agents - read the board, decide what can move, and dispatch work behind one batched approval. Use when several agent sessions are running and nobody is deciding what happens next, when dispatch requests are waiting, or when work has stalled and you need to know which item is actually blocked and on what. Triggers on "drain the outbox", "what should be dispatched", "run the board", "who is working on what", "/agent-orchestration".

learningmatter-mit Skill

Guide for selecting the most appropriate foundation MLIP model based on simulation requirements.

alpayc Skill

Answers how-do-we-do-X-here from the code and writes the answer back to docs/agent/wiki/ so it is looked up next time. Use it for a repeated question about the codebase, or after working something out that others will need.

alpayc Skill

Presents a finished or stopped run in the conversation for review — decisions first with a recommendation, screenshots rendered inline, gaps named — then applies the verdicts immediately. Use it when asked to review a run, to go through what happened overnight, or when a run has handed over. The review happens here, so nothing has to be copied between a viewer and the agent.

emanuelegreco29 Bundle

Backend engineering guidance - API design and contracts, data modeling and databases, authentication and authorization implementation (sessions vs JWT, password hashing - see codexpiator-security for the threat-model/hardening side), error handling and logging, caching, background jobs and queues, resilience and rate limiting, backend performance, and backend testing. Use for any server-side/API/data-layer question.

emanuelegreco29 Bundle

Security guidance - secure coding checklist, authn/authz threat-model and hardening (MFA, IDOR/BOLA, session fixation, password-reset abuse - see codexpiator-backend for base auth implementation), input validation and injection defense, secrets and config management, dependency and supply-chain risk, infrastructure and access control, business-logic/webhook security, AI/agent security (prompt injection, excessive AI permissions), and security testing. Use for any security-sensitive question or before shipping anything that touches auth, user input, payments, AI/agent features, or infrastructure config.

emanuelegreco29 Bundle

AI/LLM integration guidance - LLM integration patterns, prompt engineering basics, agentic and tool-use safety and security (prompt injection, excessive AI permissions, unvalidated AI output), and MCP usage and recommendations. Use when integrating an LLM, building an AI feature or agent, writing prompts, evaluating an MCP server, or assessing whether an AI/agent feature is safe.

alpayc Skill

Resolves business vocabulary to code — which class, route, endpoint and table a domain term actually lives in, and the legacy names it still goes by. Run it when a ticket's wording has no obvious home in the code, when the same concept appears under several names, or when onboarding to an unfamiliar domain. Kept in docs/agent/glossary.md.

alpayc Bundle

Surveys an unfamiliar repo and writes what it found to docs/agent/repo-map.md — workspace kind, every project with its stack, boundary rules, gate commands, changed-set command. Run it when asked what a repo contains, which projects or which stack it has, how it is tested or built; when landing in a repo not yet mapped; when the map is stale; or before any skill that reads it.

alpayc Skill

Teaches an agent the conventions, reasons and gotchas a repo never wrote down — captured as a CLAUDE.md hierarchy, root plus one per project that deviates. Run it when a repo has no agent context, when a run failed on an undocumented convention, or when a postmortem produces a fix that belongs in context.

grzmol Bundle

Salesforce security model and secure-coding enforcement for Apex, LWC, and metadata - org/object/field/record layers, permission sets vs profiles vs permission set groups vs muting permission sets, CRUD and FLS enforcement with WITH USER_MODE, AccessLevel.USER_MODE, Security.stripInaccessible, and Schema describe checks, sharing keywords and Apex managed sharing __Share records, Lightning Web Security and CSP, SOQL injection, XSS, open redirect and CSRF fixes, Named Credentials and protected custom metadata for secrets, Shield Platform Encryption impacts, and Event Monitoring. Use when writing or reviewing @AuraEnabled controllers, triggers, dynamic SOQL, lwc:dom="manual" markup, permission set metadata, integration users, or when running vf-check analyzer or the sf-security-reviewer agent.

kimsb2429 Bundle

How to handle {{S_xxxxxxxx}} secret placeholders. Use when a prompt or file contains a {{S_...}} token, or when the user asks to store, place, or use an API key, token, or password that was redacted.

stas2-git Bundle

Use when deciding which bundled agent skill to use for coding-agent workflow, agent engineering, skill design, ADK implementation, deployment, security, human review frontends, protocol selection, self-improvement, or Kaggle capstone planning.

grzmol Bundle

Build automated Salesforce UI tests - pick the cheapest tier that can catch the bug (Apex unit, LWC Jest, UTAM end-to-end, Agentforce agent tests), scaffold a UTAM plus WebdriverIO project with Salesforce's own page objects, authenticate a test run without committing a credential, and keep the suite deterministic in CI. Use when writing or repairing browser-level tests, not when configuring an org.

grzmol Bundle

Selects and implements the right Salesforce asynchronous execution model - Queueable Apex with Finalizers, Batch Apex, Apex Cursors, Schedulable Apex with CRON expressions, future methods, Platform Events, and Change Data Capture - and covers chaining limits, scope sizing, Database.Stateful, transaction boundaries, idempotency, duplicate delivery, and job monitoring or abort through AsyncApexJob and CronTrigger. Use this skill when work involves System.enqueueJob, System.attachFinalizer, Database.executeBatch, Database.getQueryLocator, Database.getCursor, System.schedule, System.scheduleBatch, System.abortJob, EventBus.publish, EventBus.RetryableException, setResumeCheckpoint, @future, BatchApexErrorEvent, ChangeEventHeader, a *__e platform event, a *ChangeEvent trigger, an AsyncApexJob or CronTrigger query, or when a long-running or high-volume process must move off the synchronous transaction.