Product & Planning

jovanipink Bundle

Apply focused PHP and Drupal engineering judgment after repository gate discovery. Use for Composer projects, Drupal modules or themes, entity and configuration APIs, caching, security, tests, compatibility, and builds; defer exact commands to repository evidence.

jovanipink Bundle

Apply focused Java and Spring engineering judgment after repository gate discovery. Use for JVM services, Spring Boot applications, concurrency, transactions, tests, dependencies, compatibility, and builds; honor repository wrappers and toolchains.

jovanipink Bundle

Apply focused C# and .NET engineering judgment after repository gate discovery. Use for .NET applications, libraries, ASP.NET Core, Entity Framework Core, async behavior, tests, packages, compatibility, and builds; defer exact commands to repository evidence.

jovanipink Bundle

Apply focused Swift and SwiftUI engineering judgment after repository gate discovery. Use for packages, applications, concurrency, UI state, tests, compatibility, and builds; require observed macOS evidence for Apple toolchain claims.

jovanipink Bundle

Apply focused PostgreSQL and SQL engineering judgment after repository gate discovery. Use for schemas, queries, transactions, migrations, indexing, compatibility, and database test concerns; never mutate shared data merely to validate.

jovanipink Bundle

Apply focused Salesforce, Apex, and Lightning Web Components engineering judgment after repository gate discovery. Use for metadata projects, packages, governor limits, data access, tests, compatibility, and deployment readiness; preserve separate live-org authority.

jovanipink Bundle

Translate technical evidence for a named stakeholder decision without overstating certainty, hiding risk, or exposing inappropriate detail. Use for executive summaries, review packets, status updates, decision briefs, and cross-functional explanations.

jovanipink Bundle

Review interoperability across agent, tool, and service protocols using exact versions, discovery records, schemas, identity, authorization, delegation, streaming, errors, retries, idempotency, cancellation, and trust zones. Use for A2A, MCP, or another agent-facing protocol boundary; use api-contract-compatibility-review for an ordinary API change without agent delegation semantics.

jovanipink Bundle

Apply focused TypeScript and JavaScript engineering judgment after repository gate discovery. Use for Node or browser packages, type safety, async behavior, tests, dependencies, compatibility, and builds; honor the repository package manager and scripts.

bostonaholic Bundle

Locates files by naming, structure, and imports. Load for read-only discovery when exact paths are unknown.

bostonaholic Bundle

Grooms a project backlog and proposes tracker changes. Trigger on "groom the backlog", "groom the board", "clean up the backlog", or "/groom-backlog"; changes require user approval.

bostonaholic Bundle

Defines pipeline artifact schemas. Load when authoring or validating artifact frontmatter, `4-repos.md`, `3-prd.md`, or design-review records.

bostonaholic Bundle

Defines when and how to write `3-prd.md`. Load during Question for vague, multi-story, cross-cutting, or behavior-replacing work.

hack23 Skill

GitHub issue creation, labeling, milestones, agent assignment, and backlog prioritization for product management

hack23 Skill

Stakeholder engagement, partnership strategies, revenue models, community building, and growth for open-source political transparency platforms

zjunlp Skill

Routing entry point for eleven families of mechanistic-interpretability methods that localize *which* internal object (layer, attention head, neuron, SAE feature, weight, or input feature) drives a model's behavior, *how influential* it is, and *what changes* when it is intervened on. Use this skill whenever the question is about a model's internal mechanism rather than its external metrics — for example, claims that a specific component is responsible for a behavior, mechanistic-evidence requests, circuit-discovery tasks, feature-attribution work (SHAP), or concept-level explanations of vision/VL models. The file lays out each family's premise, signal, cost, advantages, limitations, and how to compose them into a cheap-screen → causal-verify pipeline. Loading is hierarchical and mandatory: after picking a family from this routing file you MUST load that family's `SKILL.md`, and after picking a submethod you MUST load that submethod's `SKILL.md` — never act on the previews in this file alone.

zjunlp Skill

Macro-level strategic directions for investigating the *mechanism* behind a model behavior — the downstream half of the project's mission (mine LLM behaviors, then explain the mechanism behind them). Use once a phenomenon is observed in a model — whether already established/known or freshly mined by `/mechanism-behavior-discovery` — and the task is to choose *how* to investigate the internal cause. It is the strategy layer above the concrete method families in `/mechanism-skills`, organized around six parallel research directions — **Location**, **Causal Intervention**, **Tuning & Editing**, **Formation Tracing**, **Unit Interpretation**, **Decision Auditing** — plus how to combine them into strategies. Domain-general: it prescribes strategy, not any single model, modality, or method.

zjunlp Skill

Mine behavioral regularities in neural-network (LLM / multimodal) models — the upstream half of the project's mission (find a behavior worth explaining, then investigate the mechanism behind it). Use this skill when the task is open-ended: surface a *new* behavioral phenomenon — a candidate claim / research direction — rather than investigate an already-named mechanism. It gives strategies for choosing which behavior to probe and how to choose the data that validates it. The output is a candidate phenomenon that hands off to `/mechanism-explore` for mechanistic investigation. Domain-general: no assumption about model family, modality, or task.

zjunlp Bundle

Circuit discovery methods localize the minimal subgraph of a transformer's computational graph (a set of nodes such as attention heads and MLPs, together with the edges that connect them) that is causally responsible for a specific model behavior. Rather than scoring individual objects in isolation, these techniques recover an end-to-end mechanism by jointly identifying which components, and which information pathways between them, must be preserved to reproduce the behavior on a task distribution.

zjunlp Skill

How to choose where (and how many sites) to intervene for any operation on internal representations — activation patching, steering, CAA, DAS, SAE feature scaling, attribution patching. Use whenever the plan declares a target block / layer / site, especially when it hard-codes a single index (`target_block: 4`, `layer: 16`) without justification or copied from another paper. Covers picking the site by screening (gradient / attribution, or an activation signal) and by heuristic (mid-to-late layers usually steer best; widen to 3–5 layers if one is inert; sweep the stack at spaced intervals), and which component to target (attention / MLP / residual stream — circuit discovery studies attention heads + MLP). Prevents two symmetric failures: too few sites → downstream norm/attention denoise it → false "no effect"; too many → off-distribution collapse and the localization claim becomes untestable. Triggers include `target_block = 4`, `sliding window = 15`, `layer ∈ [12,18]`, `early blocks (0-7)`, "steering had no e

zjunlp Bundle

Use when analyzing neural network circuits, performing attribution patching, automated circuit discovery, or investigating model interpretability through edge attribution methods in transformer models

zjunlp Bundle

Use this skill when working with circuit discovery in language models, mechanistic interpretability, activation patching, attribution patching, or Layer-wise Relevance Propagation (LRP) for neural network analysis

zjunlp Bundle

Use this skill to discover and analyze computational circuits within autoregressive transformer language models via attribution methods like Edge Attribution Patching with Integrated Gradients (EAP-IG), enabling mechanistic interpretability studies and circuit evaluation.

zjunlp Bundle

Activates when working on mechanistic interpretability of transformers, specifically for automated circuit discovery in transformer models using computational graph editing and analysis.

shalomb Bundle

Use when a TODO.md backlog must be burned down by multiple agents working concurrently in isolated git worktrees. Requires both an existing TODO.md and worktree isolation — without those, use `dispatching-parallel-agents` for in-session parallel tasks, or a `*-sub-agent` skill to delegate a single task. This is the top-level orchestrator: it composes those skills into waves, handling cross-task dependency analysis, model spread across Ralph/Bart, and feedback accumulation between waves. Triggers on: 'agent-mux', 'wave of agents', 'multi-agent orchestration', 'burn down the backlog in parallel'.

shalomb Bundle

Launch a pi coding agent as a sub-process in a tmux pane, monitor its progress via JSONL streaming, and report results. Use when delegating a well-defined task (bug fix, feature, refactor) specifically to pi, with live observability. For another CLI use the matching skill (claude-sub-agent, gemini-sub-agent, copilot-sub-agent, kiro-sub-agent); for a whole TODO.md backlog across worktrees use agent-mux. Triggers include "run pi agent", "delegate to pi", "pi sub-agent", "Ralph loop in pi", or any request to run pi in a tmux pane and monitor it.

shalomb Bundle

Query and manage TargetProcess/Apptio entities (user stories, bugs, tasks, features) via tpcli. Supports discovery, filtering, pagination, entity operations, and complex team-based PI planning scenarios. Use when working with TargetProcess work items, managing project entities, or querying issue trackers for team health and status.

shalomb Bundle

Build Agent — burns down a TODO.md task list end to end, unattended. Use when there is an existing TODO.md to execute: reads it as the plan, implements each task, updates task status on completion, creates an Atomic Commit Protocol (ACP) commit per task, then progresses to the next until the list is done. Implementation uses red-green-refactor TDD with INVEST decomposition, a Farley per-test checklist, and Conventional Commits. Not for one-off changes with no task list — for those use `test-driven-development`. Triggers on: 'ralph', 'build agent', 'burn down the TODO', 'work through TODO.md', 'execute the backlog'.

shalomb Bundle

Find, cache, and refresh git repositories — both local projects and remote repos. Uses the gum CLI to discover existing checkouts under ~/*/ before cloning. Use when the user references a git repository (URL, org/repo, or local path) or when you need to locate a project on disk.

shalomb Bundle

Use when the user mentions Jira issues (e.g., "PROJ-123"), asks about tickets, wants to create/view/update issues, check sprint status, or manage their Jira workflow. Triggers on keywords like "jira", "issue", "ticket", "sprint", "backlog", or issue key patterns.

shalomb Skill

Find, create, edit, and update notes in an Obsidian vault using the `ob` CLI and direct file operations. Manages TODO lists (checking off tasks, adding items), commits changes with git, and syncs the vault. Use when the user asks to: take a note, capture an idea, update _todo.md, check off a task, find a note, add to Incoming.md, review TODOs, commit vault changes, or sync the obsidian vault. Personal notes only — for a code TODO.md backlog use `ralph-build-agent`; for formatting a code commit use `git-commit-formatter`. Triggers on: 'obsidian', 'vault', 'take a note', 'capture an idea', 'jot down', 'add to inbox', '_todo', 'my todo list', 'sync vault', 'commit vault'.

chendongqi Bundle

需求编写助手 - 编写软件需求文档。适用场景：(1) 将模糊想法转化为结构化功能需求，(2) 编写用户故事(User Story)和验收标准，(3) 创建技术规格说明书(Tech Spec)，(4) 需求优先级评估和排序，(5) 需求变更影响分析，(6) PRD/MRD/BRD等产品文档编写，(7) API接口需求定义。触发词：需求、PRD、用户故事、功能规格、技术规格、验收标准、需求文档。

chendongqi Skill

HUMAN 3.0 development assessment framework. Conducts adaptive interviews across four quadrants (Mind, Body, Spirit, Vocation), identifies consciousness levels (1.0 Conformist, 2.0 Individualist, 3.0 Synthesist), detects false transformations, assesses Glitch readiness, and provides lifestyle integration strategies. Actions: assess, interview, evaluate, guide, transform, analyze. Domains: consciousness development, personal growth, lifestyle design, problem-solving, risk assessment. Quadrants: mind, body, spirit, vocation. Levels: 1.0, 2.0, 3.0. Phases: dissonance, uncertainty, discovery. Special cases: channels, glitches, regression, false transformation. Integrates Spiral Dynamics, AQAL, ego development, flow psychology.

chendongqi Bundle

天赋与特长发现顾问，通过10个精心设计的深度问题，一次一个地引导用户探索隐藏的天赋和特长，收集具体案例，区分真正热爱与勉强学会的技能，最终给出精准的擅长领域分析和变现建议。适用场景包括 (1) 用户想要发现自己的隐藏天赋和特长，(2) 需要明确职业发展方向和优势领域，(3) 寻找个人特长的变现路径，(4) 区分兴趣爱好与真正擅长的领域。触发关键词为天赋发现、特长挖掘、优势分析、能力评估、职业定位、变现建议。

somtougeh Skill

Execute an implementation plan or PRD through complete, verified stories, preserving dependencies and progress across sessions.

drolu Bundle

Agentless asset discovery and CMDB on RTerm — discover Windows (WinRM/PowerShell), Linux/UNIX (SSH), network devices (SSH/SNMP), VMware vCenter/ESXi, and databases (MS SQL/Oracle/MySQL/Sybase), consolidate zone results outbound-only via the WebSocket gateway or NATS mesh, and maintain a versioned, diffable, queryable inventory. Covers the full BMC Helix Discovery (ADDM) protocol matrix using RTerm's backends, playbooks, and inventory ledger. Use when an agent needs to collect infrastructure inventory, build/maintain a CMDB, run discovery scans, or reconcile assets.