agentleFS
jwrobesjwrobes/skills/continue-workflow/SKILL.md

continue-workflow

Skill1 starsChanged 7 months agoMITView on GitHub

Resume a previous working session by finding and loading the most recent continuation prompt. Use when the user says "continue", "pick up where I left off", "resume", "what was I working on", or starts a new session in a project workspace or docs folder and wants to restore context.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 8 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Search for continuation prompts and checkpoints in the current directory tree and common locations.

    line 19
  2. Distinctive:

    When prompts are found, rank them and present to the user:

    line 47
  3. Very specific:

    Folder match first: If the current directory is inside a specific initiative folder (e.g., workbench/local_judge_parallelization/), the …

    line 49
  4. Distinctive:

    After reading all context files, present a brief summary:

    line 84
  5. Distinctive:

    Before starting work, do a quick consistency check:

    line 105
  6. Distinctive:

    If in a project workspace:

    line 113
  7. Very specific:

    If no CONTINUATION_PROMPT*.md or CHECKPOINT.md exists:

    line 142
  8. Distinctive:

    Read the README for the current workspace/project

    line 146
  9. Distinctive:

    Check recent git history: git log --oneline -10

    line 147
  10. Distinctive:

    Ask: "Would you like me to read the README and recent history to figure out where things stand?"

    line 149
  11. Distinctive:

    Respect the "Key Decisions Made" section: Don't re-litigate settled decisions unless the user explicitly wants to revisit them.

    line 154

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Continue Workflow

Find and load the most recent continuation prompt to resume work from a previous session.

When to Use

  • User starts a new session and says "continue", "resume", "pick up where I left off"
  • User opens a workspace and wants to know the current state of an initiative
  • User says "what was I working on?"

Step 1: Find the Continuation Prompt

Search for continuation prompts and checkpoints in the current directory tree and common locations. Continuation prompts use descriptive filenames: CONTINUATION_PROMPT--<slug>.md (the slug describes the session state). Legacy files named exactly CONTINUATION_PROMPT.md are also matched.

Determine search scope from the current working directory:

If in a project workspace (*_workspace/):

# Use glob to find all continuation prompts in the workbench
find ~/workspace/<project>_workspace/workbench/ \
  \( -name "CONTINUATION_PROMPT*.md" -o -name "CHECKPOINT.md" \) \
  2>/dev/null

If in a documentation repo (has a master index file like *-INDEX.md):

find <docs_repo_root>/ -maxdepth 3 \
  \( -name "CONTINUATION_PROMPT*.md" -o -name "CHECKPOINT.md" \) \
  2>/dev/null

If the location is unclear, search the workspace root:

find ~/workspace/ -maxdepth 5 \
  \( -name "CONTINUATION_PROMPT*.md" -o -name "CHECKPOINT.md" \) \
  2>/dev/null

Suggest the best match, then ask

When prompts are found, rank them and present to the user:

  1. Folder match first: If the current directory is inside a specific initiative folder (e.g., workbench/local_judge_parallelization/), the prompt from that folder is the top suggestion.
  2. Recency second: Sort remaining prompts by last-modified date (newest first).
  3. Extract the slug from the filename to display a human-readable description. The slug is everything between CONTINUATION_PROMPT-- and .md, with hyphens replaced by spaces.
Found continuation prompts:

→ 1. local_judge_parallelization/CONTINUATION_PROMPT--judge-concurrency-plumbed-run-5k-baseline-next.md
     "judge concurrency plumbed, run 5k baseline next" (2026-04-10 08:05) ← current folder
  2. sns_sqs_architecture/CONTINUATION_PROMPT--event-schema-v2-migration-next.md
     "event schema v2, migration next" (2026-04-09 16:30)
  3. coverage_transfer_bug/CONTINUATION_PROMPT--root-cause-found-fix-pending.md
     "root cause found, fix pending" (2026-04-07 11:15)

Which would you like to continue? [1]

Get last-modified dates with: stat -f "%Sm" -t "%Y-%m-%d %H:%M" <file> (macOS)

Always ask the user even if there's only one match — confirm before loading context.

Priority order when multiple prompts exist in the same directory:

  1. CHECKPOINT.md (if newer than any CONTINUATION_PROMPT*.md) — represents auto-saved state from a more recent session
  2. The most recent CONTINUATION_PROMPT*.md — represents an intentional close-out

Step 2: Load and Execute the Prompt

  1. Read the continuation prompt file
  2. Read all files listed in the "Context Files to Read" section, in the order specified — this is critical for rebuilding context
  3. Review the "Current State" and "Open TODOs / Next Steps" sections
  4. Note the "Key Decisions Made" and "Key Constraints" sections — these are guardrails for the session

Step 3: Present Summary to User

After reading all context files, present a brief summary:

Resuming: <Initiative/Project Name>
Last session: <date from the prompt>

What was done:
- <bullet from Session Summary>
- <bullet from Session Summary>

Current state: <1-2 sentences from Current State section>

Next steps:
1. <top TODO>
2. <second TODO>

Ready to continue with "<top TODO>", or would you like to do something else?

Step 4: Verify State

Before starting work, do a quick consistency check:

# Check for uncommitted changes (might have changed since the prompt was written)
cd <workspace_or_docs_root>
git status --short

If in a project workspace:

# Verify the worktree still exists
ls -d <worktree_path> 2>/dev/null

# Check worktree status
cd <worktree_path>
git status --short

If there are unexpected changes (files modified since the continuation prompt was written), flag them:

Note: These files have changed since the last session close-out:
- <file> (modified after the continuation prompt was written)
You may want to review these before continuing.

If the worktree no longer exists (was cleaned up), tell the user:

The worktree at <path> no longer exists. The branch <branch_name> may still exist in the project repo.
Want to re-create the worktree, or has this initiative been completed?

Step 5: Begin Work

Once the user confirms what to work on, proceed with the task. Keep the continuation prompt's constraints and decisions in mind throughout the session.

When No Continuation Prompt Is Found

If no CONTINUATION_PROMPT*.md or CHECKPOINT.md exists:

  1. Tell the user: "No continuation prompt found in the current directory tree."
  2. Offer to help reconstruct context:
    • Read the README for the current workspace/project
    • Check recent git history: git log --oneline -10
    • List recently modified files: find . -name "*.md" -newer <some_reference> -type f
  3. Ask: "Would you like me to read the README and recent history to figure out where things stand?"

Guidelines

  1. Always read all context files. The continuation prompt was written specifically to restore context — skipping files means working with incomplete information.
  2. Respect the "Key Decisions Made" section. Don't re-litigate settled decisions unless the user explicitly wants to revisit them.
  3. CHECKPOINT.md takes precedence over CONTINUATION_PROMPT*.md if both exist in the same directory and CHECKPOINT.md has a newer timestamp — it represents a more recent auto-saved state.
  4. Don't modify the continuation prompt during this skill. It's a read-only input. A new one will be written by /close-out-session at the end of the session.
  5. If the continuation prompt references files that no longer exist, flag this to the user rather than silently skipping them — it might indicate work was completed or reorganized.

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.