Context Save

Context Save

Capture enough state that a cold read of the saved file is sufficient to resume work with no other memory.

Steps

1. Gather git state:

   git rev-parse --abbrev-ref HEAD
   git status --short
   git diff --stat
   git log --oneline -10
   

2. Determine the save location: .agent-context/ at the repo root. Create it if missing, and make sure it's covered by .gitignore (add a .agent-context/ line if not already ignored) — this is working scratch state, not something to commit.

3. Write a new file .agent-context/<UTC timestamp>-<slug-of-title>.md (never overwrite an existing save; each save is a new file). Infer a short title (3-6 words) from the conversation if the user didn't give one.

   File contents:

   ---
   branch: <current branch>
   timestamp: <ISO-8601>
   files_modified:
     - <path>
   ---
   
   ## Working on: <title>
   
   ### Summary
   <1-3 sentences: the goal and current progress>
   
   ### Decisions made
   - <architectural choice or trade-off, and why>
   
   ### Remaining work
   1. <next step, in priority order>
   
   ### Notes
   <gotchas, things tried that didn't work, open questions>
   

4. Confirm to the user with the file path and a one-line summary of what was saved.

Rules

• Never modify code — this skill only reads state and writes the context file.
• Always include the branch name — needed for cross-branch resume via context-restore.
• Infer content from git state and conversation history; only ask the user if the title or summary genuinely can't be inferred.