Bootstrap

Skill jerlifish/slimecontext/skills/bootstrap

Resume work with the most relevant saved context. Lists recent slimecontext sessions, lets the user pick a topic, and loads that topic's context chain (project → feature → sub-feature) into the conversation. Use when the user runs /slime:bootstrap or asks "what was I working on", "resume where I left off", "load my project context", "what were the next steps", or starts a session clearly continuing earlier work. Their saved context lives in the slimecontext store outside the repo and is only reachable through this skill — consult it before exploring the repo to answer "where was I" questions.From its SKILL.md

Install

npx -y skills add jerlifish/slimecontext --skill bootstrap

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.1 KB, as published. Nobody here has run it

slime:bootstrap

Load exactly the context the user needs to continue — no more. Read only the small metadata file here; delegate note-reading to a subagent so intermediate file dumps stay out of this conversation.

1. Resolve the notes root

cat ~/.claude/slime.json → root. Missing file or missing root directory → slimecontext isn't set up: say so and offer to run ${CLAUDE_PLUGIN_ROOT}/skills/setup/SKILL.md now. Stop if declined.

2. Pick a topic

cat <root>/.slimecontext — this is the ONLY file to read directly; do not open any context.md yourself.

Rank recent_sessions for relevance: if the working directory or git remote get-url origin matches a project's repos, that project's sessions come first; within a project, newest first. No match → all sessions, newest first.

AskUserQuestion with up to 4 options — label = topic, description = when · level · summary. If more sessions exist than fit, pick the 4 most relevant and mention the user can name any other project/feature via "Other" (match free-text answers against nodes paths and tags).

If .slimecontext has no sessions at all, say the notes are empty and suggest /slime:save at the end of this session instead.

3. Assemble the context chain (subagent)

Launch one general-purpose Agent (synchronous) — it can't see this conversation, so give it full paths:

Notes root: <root>
Chosen node: <path, e.g. myapp/auth/oauth-refresh>

Read the context.md of the chosen node AND of each ancestor
(<root>/myapp/context.md, <root>/myapp/auth/context.md, …), ordered
project → feature → sub-feature.

Return them as one bundle, per note: a "# <path>" top-level header (the
notes' own sections are "##", so note boundaries must outrank them), then the
note's Now, Decisions, and Direction sections in full, and only the 3 newest
Log entries. No commentary, no reformatting — the notes are already condensed.

If the Agent tool isn't available in this context, assemble the bundle yourself by the same rules — read only the chain's context.md files, nothing else.

4. Resume

Present the bundle's essence, then get to work:

"Resuming <topic> (<project>)."
Where things stood (from the deepest note's Now).
"Next steps were: …" (the open - [ ] items).

Keep the full bundle in mind as working context — that's the point of loading it. Don't re-summarize it back to the user at length; they wrote these notes.