Personal skills for kevmoo - but shared with anyone why is interested!
Overview
This is my personal repository for specialized . These skills provide procedural, actionable instructions to AI assistants for specific domains or tasks.
README
This is my personal repository for specialized Agent Skills. These skills provide procedural, actionable instructions to AI assistants for specific domains or tasks.
Skills Inventory
To install any skill individually:
dart run skills@ add kevmoo/kevmoo_skills --skill
# or
npx skills add kevmoo/kevmoo_skills --skill
| Skill | Description | Key Features |
|---|---|---|
| clarify-confirm-continue | Orchestrates a disciplined task intake workflow across code, docs, and project files by autonomously fact-checking the workspace, resolving genuine decision forks, and confirming execution readiness via ask_question before making modifications. Use when starting multi-step tasks, refactorings, ambiguous feature requests, broad documentation updates, or when invoked via /clarify-confirm-continue or “ccc”. Don’t use for trivial single-turn lookups, straightforward single-line edits where intent is unambiguous, or emergency rollbacks. | |
| cold-read | Reads a draft issue, pull request, bug report, commit message, or review comment as a busy, fresh-context reader and reports whether that reader can make their next decision from the first few lines. Use when invoking /cold-read, before posting any agent-written artifact to a human audience, or when measuring how quickly past artifacts let a triager, code owner, or reviewer decide. Don’t use for reviewing code diffs (use pr-review), triaging existing review comments (use pr-triage), or copy-editing prose style. | Fresh-context reader simulation (triager, owner, reviewer), Decision line and words-above-fold measurement, Known-to-reader and cut-list detection, Machine-readable JSON verdict for gating and batch baselines |
| dart-cleanup | Orchestrates specialized Dart refactoring, code quality, testing, and modern language features on demand. Use when cleaning up, modernizing, refactoring, or optimizing Dart code. Don’t use for non-Dart projects or general codebase search. | |
| distilling-strategies-interactively | Analyzes unstructured artifacts (proposals, design docs, issues, notes) to identify gaps, contradictions, and implicit assumptions, and interviews the user via a structured quiz to distill a definitive strategy and trade-off matrix. Use when resolving competing architectural proposals, clarifying ambiguous project goals, identifying blind spots in early design documents, or synthesizing consensus from fragmented team notes. Don’t use for simple text summarization without strategic synthesis, writing code solutions directly, or managing task lists without strategic decision-making. | Artifact mapping & contradiction triage, Socratic interviewing & interactive quiz, Scenario modeling & trade-off matrix, Strategy synthesis & action plan |
| duckdb | Queries, inspects, transforms, and converts local structured and semi-structured data files (JSON, JSONL, NDJSON, CSV, TSV, Parquet, SQLite), unified code diffs, and Markdown documents on Linux and macOS with DuckDB SQL. Use when querying local JSON, CSV, or SQLite files, benchmark artifacts, large code diffs, or Markdown corpora, inspecting schemas and row counts, searching and joining JSONL log globs with SQLite index databases (such as transcript.jsonl and conversation_summaries.db), joining files across formats, converting datasets to Parquet, or filtering stdin pipelines. Prefer it over ad hoc Python or jq for filtering, grouping, and joining local data. Don’t use for remote cloud data warehouses (BigQuery, Snowflake, Spanner) or PLx dashboards. | Non-interactive subshell flag safety (-batch -dark-mode), Fast JSONL, CSV, Parquet, and SQLite querying, Explicit schema projection & multi-file globbing, Cross-format SQL joins & Parquet materialization, Relational code diff querying (scripts/diff_macros.sql), Relational Markdown indexing (scripts/markdown_macros.sql) |
| gerrit-stacked-cls | Best practices for managing stacked changelists (CLs) in Gerrit using depot_tools, avoiding common pitfalls with Change-Ids. | Stacked CLs, Gerrit, depot_tools, Change-Ids |
| just-brainstorm | Ponder architectural designs, explore potential solutions, weigh tradeoffs, or analyze the impact of proposed codebase changes without making code changes. Always outputs the analysis to a Markdown artifact. | Architectural design, Trade-off evaluation, Impact & “what-if” analysis, Codebase metric gathering |
| markdown-long-lines | Ensure markdown prose lines wrap within 80 columns, excluding code blocks, URLs, and tables. | Markdown formatting, 80-column line wrapping, Formatting exclusions |
| new-worktree | Creates and initializes a new Git worktree as a sibling directory with automated branching and naming rules based on the latest remote default branch. Use when invoking /new-worktree or requesting a new Git worktree for parallel development, feature work, or bug fixes in external repositories under ~/github. Don’t use for the Dart SDK (~/github/dart-sdk), non-Git repositories, or repositories outside ~/github. | Git worktree setup off latest origin/main or origin/master, Sibling directory placement, Hard boundary checks for non-Git, outside ~/github, and Dart SDK, Automated branch and folder naming |
| pr-review | Reviews GitHub Pull Requests or local Git branch diffs for defects, removed behavior, and specification drift using a 13-angle rubric and adversarial noise filtering. Use when reviewing a GitHub PR (#N or URL), auditing a local feature branch before opening a PR, or invoked via /pr-review. Don’t use for triaging existing PR review comments or CI failures (use pr-triage), multi-repo branch cleanup sweeps (use pr-cleanup), or non-GitHub code review systems. | Explicit 3D Paranoia Tiering (Ring, Confidence, One-Way/Two-Way Door) with --deep escalation, Two-Axis Evaluation (Linked Issue Spec Traceability + RUBRIC.md Standards), 13 analytical review angles + Inquisitor Presumption of Theater doctrine, 2-Tier ownership action gate (own repo auto-fix vs. external read-only) |
| quick-question | Provide ultra-concise, direct technical answers and rapid clarifications without heavy overhead, deep research loops, or artifact generation. (Triggers on “quick question” or “qq”) | Concise Q&A, rapid clarification, Inline answers (no artifacts), Read-only guidance |
| quiz-me | Builds a primary-source-grounded prerequisite graph (or linear mastery track) of atomic concepts and tests the user across unfamiliar technical domains, specifications, or codebases one concept at a time with hard comprehension gates, targeted teaching blocks, and fresh transfer questions. Use when invoking /quiz-me, “quiz me”, verifying that you truly know and have internalized a complex domain or architecture from first principles, or mapping prerequisite concepts and key mental shifts with active comprehension checks. Don’t use for single-turn factual lookups (use quick-question), unquizzed learning walkthroughs, document/code review, co-editing, or issue triage (use slice-and-dice), branching design decision trees (use grilling or clarify-confirm-continue), or passive text summarization. | Hard comprehension gates (“SND + retrieval practice”) across a prerequisite DAG or linear checklist, Open-ended or inline “Pick + Explain Why” scenario checks with anti-guessing calibration prompts, Root-first default with conversational skip-ahead and immediate teaching on “teach me” or any miss, Strictly 1 active concept per turn with generous plain-English grading and zero robotic headers, Progressive disclosure via references/graph_schemas.md and references/miss-diagnosis-examples.md |
| sem-cli | Use the sem CLI (sem-cli) for fast local code exploration, instant call-graph navigation (callers/refs), signature-only context packing (--headers), hotspot and co-change discovery (sem log), transitive impact analysis (sem impact), and entity-level semantic diffs (sem diff). |
Instant call-graph navigation, signature-only context packing, entity-addressed substring search, hotspot and co-change analysis, transitive impact analysis, semantic diffs |
| sharpen-later | Bookmarks agent workflow friction (a failing CLI, an ignored rule, a missing skill or hook) in one command, then resumes the active task in the same turn. Use when the user invokes /sharpen-later, says “log this for sharpen-saw”, or wants to note a papercut without derailing current work. Don’t use for root-cause analysis or for fixing tools and rules now (use sharpen-saw). | One-command friction bookmark, Same-turn task resumption, Automatic session and timestamp capture, Ledger shared with sharpen-saw |
| sharpen-saw | Audits past Claude Code session transcripts offline to find recurring agent friction (failing commands, edit spirals, permission denials, context hogs, dormant skills) and turns the findings into durable fixes across rules, hooks, CLI shims and skills. Use when the user asks to “sharpen the saw”, audit past sessions for friction, fix repetitive agent mistakes, drain the /sharpen-later queue, or runs /sharpen-saw. Don’t use for debugging the feature currently being built, or for logging a papercut mid-task (use sharpen-later). | Single-pass offline transcript scan, Error clusters with succeeding-sibling contrast, Rule and skill budget audit, Multi-tier fixes (rules, hooks, shims, skills), Drains the /sharpen-later queue |
| sidequest | Synthesizes conversation history and active tasks into a visual hierarchy map (sidequest.md) backed by a deterministic JSON state file (sidequest.json). Supports multiple sequential and concurrent main quests, sub-quests, and side-quests with automatic hierarchical numbering and completion sequencing. Use when the user invokes /sidequest, asks where we are, what we were doing, or what’s on our stack, or when the conversation branches across multiple topics, blockers, or digressions. Don’t use for simple one-off questions. |
Conversation mapping, Task hierarchy & numbering, VCS state tracking, Subagent history audits |
| slice-and-dice | Walks through multi-part documents, proposals, technical audits, issue backlogs, or codebases in bite-sized (~60-second) slices ([SND 1/N] … [SND N/N]) with a persistent progress tracker, auditable Source anchors, two-tier transition footers, and dynamic plan updates. Use when learning or understanding material without being quizzed, reviewing, co-editing, or triaging a dense document, RFC, plan, or audit slice-by-slice without cognitive overload, or when invoked via /slice-and-dice, /slice-n-dice, /snd, “slice and dice”, or “slice-n-dice”. Don’t use for single-turn answers (use quick-question), active mastery quizzing with hard comprehension gates (use quiz-me), branching design decision trees (use grilling or clarify-confirm-continue), or one-shot automated PR diff reports (use pr-review). | Bite-sized ~60-second slices ([SND 1/N]) for unquizzed learning, review, and co-editing, Pre-flight scope gate (N > 10) and hard boundary against branching decision trees, Two-tier action footers preventing the “next, next, next” suggestion-skipping trap, 4 clean tracker states ([ ], [-], ⏳ Deferred, ☑️ Done) with zero “skip”/“drop” ambiguity, Dynamic plan evolution with explicit upfront realization callouts |
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add kevmoo/kevmoo_skills