A-NGJ

Go from a task description or research note straight to a living behavioral spec plus a goal envelope — a ready-to-run work order for a condition-based agent loop — in one pass, with no separate design to review and no phased plan. Use when user says 'make a goal-ready spec', 'spec this for an agent loop', 'turn this into a goal for /goal', or 'one-shot a spec and goal from this'. Do NOT invoke for genuine tradeoffs the user wants to review (use rpi-propose), a narrow scoped change wanting a phased plan (use rpi-plan), or a phased plan without a design (use rpi-blueprint).

Move completed or superseded .rpi/ artifacts to .rpi/archive/ with reference checks and frontmatter updates. Use when user says 'archive these artifacts', 'clean up .rpi/', or 'archive completed work'.

Go from a research note or a short problem statement straight to a phased plan in one pass, without a separate design to review. Use when user says 'just get me to a plan', 'skip the design, plan it', 'one-shot this from the research note', or 'I don't need a design, just plan this'. Do NOT invoke for genuine tradeoffs, multiple defensible approaches, or high-blast-radius changes — use rpi-propose instead; for a narrow scoped change to existing behavior use rpi-plan.

Create git commits for the current working tree with user approval, sensitive-content scan, and gitignore enforcement. Use when user says 'commit this', 'commit the changes', 'make a commit', or 'let's commit what we have'.

Diagnose a bug iteratively: reproduce, find root cause, attempt up to 3 fixes, produce a diagnosis artifact. Use when user says 'why is X not working?', 'X is broken', 'after Y, I see Z fail', or reports broken behavior, even if they don't say 'diagnose'. Do NOT invoke for understanding working code (use rpi-research).

Walk through a recent diff and explain what changed and why, focusing on non-obvious decisions. Use when user says 'explain the changes', 'walk me through the diff', 'what did we just do?', or 'explain how <feature> was implemented'.

Capture in-flight conversation context to a deterministic per-project temp file at the end of a session, so the next session in the same project can be told to read and consume it. Use when user says 'handoff', 'end of session', 'save session context for next time', 'wrap this up for the next session', or 'hand off to the next session'.

Execute an approved plan from .rpi/plans phase by phase with TDD and per-phase verification. Use when user says 'implement the plan', 'start implementing', 'run the next phase', or just approved a plan.

Read, search, create, and edit notes in the user's personal Obsidian vault, located at "$HOME/Documents/Remote vault primary". Use this skill WHENEVER the user refers to "my notes", "my vault", "my Obsidian", a "daily note", or asks to find / look up / write down / jot / capture / file something into their notes — even if they don't say the word "Obsidian". Also use it when the user asks to summarize, link, or reorganize existing notes, or to add to today's daily note. If a request plausibly concerns the user's own knowledge base rather than the wider web, reach for this skill before falling back to generic file tools, so that Obsidian conventions (wikilinks, tags, frontmatter, daily notes) are respected.

Compact reference of best practices for authoring Agent Skills (SKILL.md files), distilled from Anthropic's official skill-authoring guide. Use whenever writing a new skill, reviewing or improving an existing SKILL.md, deciding how to structure a skill's files, or when the user asks about skill conventions, descriptions, progressive disclosure, or skill quality — even if they don't say "best practices".

Turn a task or research note into a living spec plus a goal envelope — a ready-to-run work order for a condition-based agent loop.

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

Complete reference for Neovim's vim.lsp API — all functions, arguments, and submodules. Use this skill whenever the user asks about vim.lsp functions, LSP keybindings, LSP configuration, or needs help with any Neovim LSP command. Also use when the user mentions lsp.buf, codelens, inlay hints, semantic tokens, or LSP-related Neovim setup, even if they don't say "vim.lsp" explicitly.

Complete reference for Neovim's vim.treesitter API — all functions, arguments, and submodules. Use this skill whenever the user asks about treesitter functions, treesitter queries, syntax highlighting, TSNode manipulation, or tree-sitter configuration in Neovim, even if they don't say "vim.treesitter" explicitly.

Answer questions from the user's configured Obsidian LLM Wiki vault, grounded only in the wiki's own pages, with interoperable Obsidian [[wiki-link]] citations and a ## References section. Use when the user asks to query or ask their wiki / knowledge base — e.g. "ask my wiki about X", "what does my wiki say about …", "query the wiki about …", "search my LLM wiki for …". Do NOT use for general questions that are not about the user's configured wiki; this skill only ever answers from on-disk wiki content and never writes to the vault.