From memex
Workflow for repositories tracked with memex — a CLI that records development as a versioned DAG of nodes under .memex/. Use whenever you are about to make a code change in a repo containing a .memex/ directory, when the user references memex, nodes, or tracked work, or when asked to create, resolve, abandon, or edit a node. Covers orienting in the graph via `memex context`, finding the right parent, branching, creating a node before writing code, recording decisions/artifacts/open-threads/rejected alternatives as you go, and resolving cleanly.
How this skill is triggered — by the user, by Claude, or both
Slash command
/memex:memexThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
memex tracks development as a versioned DAG of nodes stored under `.memex/`. Every feature, fix, or refactor gets a node. This skill is the contract for working in a memex-enabled repository.
memex tracks development as a versioned DAG of nodes stored under .memex/. Every feature, fix, or refactor gets a node. This skill is the contract for working in a memex-enabled repository.
A repository is memex-enabled if .memex/ exists at the repo root. If absent, this skill does not apply — ask the user whether to run memex init before proceeding.
Confirm the CLI is on PATH:
memex --version
If the command is missing, stop and tell the user — do not try to write to .memex/ by hand.
Whenever you are joining in-progress work, resuming after a break, or evaluating a candidate parent for new work, start with:
memex context --format xml
This returns the project root's goal/decisions, the trimmed ancestor chain, and the active node's full summary in a single structured payload. Prefer this over chaining memex node show + memex graph — one call, output shaped for parsing.
memex context --id <id> --format xml to orient on a specific node instead of the active one.--format xml for agent consumption (most reliable to parse); --format markdown if the user will read it.--depth N to widen the ancestor window on deep branches; --depth 0 for a minimal root+current payload.Find the parent. Choose based on what your work depends on, not what is most recent.
memex graph view — visualize the DAG.memex node list — IDs, parents, statuses, goals.memex search <keyword> — full-text search over node summaries; use domain terms (e.g. config, search, rename).memex context --id <candidate> --format xml to confirm its scope matches what you're about to attach to.Branch from main: git checkout -b <type>/<name> (feat/..., fix/..., chore/..., test/..., docs/...).
Create the node before writing code:
memex node create --parent <id> --goal "<goal>"
A short placeholder goal is fine if scope is uncertain — refine it later with memex node edit --goal "...".
Record context as you go, not after. Each flag appends without touching other fields (except --goal, which overwrites):
memex node edit --decision "chose X over Y because Z"
memex node edit --artifact "path/to/key/file"
memex node edit --open-thread "question to revisit"
memex node edit --rejected $'description = "Alternative approach"\nreason = "Why rejected"'
memex node edit --goal "Updated goal if scope changed"
Use --summary only for a full bulk replacement (e.g. bootstrapping from a plan).
Two prompts to keep in mind:
--decision you record, ask: what alternative did I deliberately not take, and why? If there's an answer, that's a --rejected.--open-thread.--rejected and --open-thread are the under-used fields. Using them is the difference between a node that captures why the work landed this way and one that only captures what shipped.
memex node resolve when the work is complete.memex node abandon if the approach turned out wrong or was superseded — leave a note in the summary explaining why..memex/nodes/ together — they describe the same unit of work..memex/state.json. It is per-developer active-node state and creates guaranteed merge conflicts. The default .memex/.gitignore already excludes it; do not override..memex/ or bundle multiple nodes per file. Always use the CLI.Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub maxrios/memex --plugin memex