agentleFS
Sign inStart free
jggomezjggomez/expert-ai-developer-skills/skills/context-restore/SKILL.md

context-restore

Skill30 starsChanged 4 days agoApache-2.0View on GitHub

Use at session start (State 0) or when asked to "load context", "catch up", or "what was done before" — prioritizes LLM Wiki Progressive Disclosure (context/memory/MEMORY.md) across Claude Code and Google Antigravity while supporting user-confirmed restore for scripted sessions (context/sessions/).

We split this skill into the rules it gives an agent

5 rules. 0 are common, 0 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 17 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Orchestrator State 0 Silent Restore: When context/memory/MEMORY.md exists, silently absorb cross-AI handoff state (<= 80-100 lines) left by …

    line 13
  2. Very specific:

    Read Master Index Only (~500 tokens): If context/memory/MEMORY.md exists, read ONLY context/memory/MEMORY.md (<= 80-100 lines) to absorb …

    line 23
  3. Very specific:

    If exists: false, continue normally without noise.

    line 31
  4. Very specific:

    Claude Code: Read context/memory/MEMORY.md silently in State 0.

    line 47
  5. Very specific:

    Antigravity: Read context/memory/MEMORY.md silently in State 0.

    line 48

“Common” is measured: kept by at least 6 in 10 of the 17 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Context Restore Skill

Overview

This skill restores durable repository memory, decisions, conventions, and active handoff state across Claude Code and Google Antigravity. It prioritizes LLM Wiki Progressive Disclosure (context/memory/MEMORY.md) for ~500-token State 0 handoffs while preserving interactive, user-confirmed restoration for standalone scripted session stores (context/sessions/).

When to Use

Trigger Scenarios

  • Orchestrator State 0 Silent Restore: When context/memory/MEMORY.md exists, silently absorb cross-AI handoff state (<= 80-100 lines) left by Claude Code or Antigravity.
  • Explicit User Handoff Resume: "load context", "catch me up", "resume where Claude Code / Antigravity left off".
  • Standalone Session Discovery: Inspecting historical archives in context/sessions/.

When NOT to Use

  • Repositories without context/memory/ or context/, or when the user explicitly asks to start fresh.
  • Saving new milestones or handoffs: Route to context-capture.

Process

Mode A: LLM Wiki Progressive Disclosure (context/memory/ — Primary)

  1. Read Master Index Only (~500 tokens): If context/memory/MEMORY.md exists, read ONLY context/memory/MEMORY.md (<= 80-100 lines) to absorb active host, branch/task, recent decisions, and pointers to root AGENTS.md (<= 90 lines), DESIGN.md (<= 110 lines), and the last 5 daily logs.
  2. Drill Down Strictly on Demand: Open context/memory/architecture.md, context/memory/conventions.md, DESIGN.md, or a specific context/memory/log/YYYY-MM-DD.md only when the active task touches that domain.

Mode B: Interactive Restore for Scripted Sessions (context/sessions/)

  1. Discover Sessions Non-Invasively:
    python3 ./skills/context-restore/scripts/context_list.py --json
    If exists: false, continue normally without noise.
  2. Rank & Ask User Before Loading: Rank recent, matching, or in-progress/blocked sessions first, then ask the user to select Full (summary.md, decisions.md, architecture.md, preferences.md), Light (architecture.md + preferences.md), Just list, or Skip (honoring restore.askBeforeLoad in context/.contextrc.json).
  3. Ingest Approved Scope & State Adoption: Unpack flows only if needed (python3 ./skills/context-capture/scripts/context_pack.py --unpack <session-id>), then state adopted decisions, resumed TODOs, and any conflicts with the current prompt.

Usage

Commands & Progressive Disclosure Reads

# Mode A — Read master wiki index first; open linked pages only on demand
cat context/memory/MEMORY.md

# Mode B — List scripted sessions (JSON or human-readable)
python3 ./skills/context-restore/scripts/context_list.py --json
python3 ./skills/context-restore/scripts/context_list.py

Host Execution Instructions

  • Claude Code: Read context/memory/MEMORY.md silently in State 0; in standalone mode, run context_list.py --json via Bash and ask (Full / Light / Just list / Skip).
  • Antigravity: Read context/memory/MEMORY.md silently in State 0; in standalone mode, use context-keeper or context_list.py --json with user confirmation.

Red Flags

  • Eager-reading all context/memory/log/ files or unpacking archives upfront instead of reading only MEMORY.md first.
  • Loading context/sessions/ archives in standalone mode without asking the user (Full / Light / Just list / Skip).
  • Silently following an outdated decision that conflicts with explicit new user instructions.

Verification

  • In State 0, only context/memory/MEMORY.md (~500 tokens) is read initially; linked pages load strictly on demand.
  • In standalone context/sessions/ mode, discovery runs cleanly and user confirms (Full / Light / Just list / Skip) before ingestion:
    python3 ./skills/context-restore/scripts/context_list.py --json
  • Active branch, decisions, and pending items are honored without token bloat.

References

  • Restore checklist

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free