AI Agent Workflow Collection

Loads relevant entries from memory/ into the conversation context. Use at the start of any non-trivial conversation, and any time the topic might overlap with what is stored in memory (personal facts, habits, goals, relationships, work, health, hobbies, etc.). Hybrid strategy — always read memory/index.md and memory/tags.md as the map first, then pull full entry bodies by tag match and by semantic match on titles. No artificial cap on how many entries can be loaded.

Orchestrator for writes into memory. Analyzes the source (the current conversation OR an external file/text passed as an argument), decides which new facts, changes, contradictions, and deprecations should be recorded in memory/, builds a diff proposal, and confirms it with the user. Only after an explicit "yes" does it call memory-write to apply changes. Triggered four ways — (1) explicit user command ("sum it up", "write down what you learned", "update memory", "/memory-retro"); (2) the agent's own initiative when meaningful new facts or contradictions surface in the dialog; (3) mandatory auto-run from grill-me — after a grill session completes, memory-retro always runs without re-asking the user; (4) bootstrap mode for bulk import from an external file (chat history dump, personal notes, session transcript) — fully autonomous, no dialog confirmations, all edge cases resolved by deterministic heuristics with confidence demoted to low as a signal for later review. In normal mode, never writes to memory silently.

Fast path for adding a movie to the watch-later wishlist. Triggers on short phrases like "add the movie X", "add to films to watch X", "in the film wishlist Y", "new movie to watch — Z", "throw the movie W into films". Before writing, it researches IMDb rating, release year, and genre via WebSearch and stores them next to the title in lists/movies.md. Not used for showing the list, marking watched, or removing — those operations go through the regular lists skill.

Shows today's schedule as a table. Synthesizes the reference schedule from lists/schedule.md and today's day card from days/YYYY-MM-DD.md. Triggered by questions like "what's my schedule?", "what should I do today?", "show me my schedule", "what's on for today?".

Creates a day check-in — a Markdown file with today's prompts in days/. Triggered by "/new-day", "new day", "let's start the day". Works in two stages — first it creates a blank template (planning horizons, abstinence, errand of the day) and stops so the user can fill it in by hand; once the user signals "done / filled", it runs a defensive cross-check against the dynamics of the last days (flags focus drift, project switching, defensive scripts being acted out), then a retrospective of the contents, optionally proposes memory updates, and commits everything in a single commit.

Service skill that picks tags for a memory entry. Invoked EXCLUSIVELY from memory-write. Guards the tag dictionary against logical and lexical duplicates (psychology / psychotherapy, work / career, sport / training, etc.). ALWAYS reads memory/tags.md in full first, then picks 2–5 existing tags. If a new tag is needed, it asks the user explicitly with a justification of why existing tags do not fit and a definition of the new one. A new tag is created only after an explicit "yes".