catchup

Use after /clear, /compact, session resume, or context loss. Use when branch context is stale or unknown. Use when starting work in an existing worktree.

INSTALLATION
npx skills add https://github.com/camacho/ai-skills --skill catchup
Run in your project or agent environment. Adjust flags if your CLI version differs.

SKILL.md

Catchup — Context Reconstruction

This skill is strictly read-only. Do NOT edit files, create branches, or run write commands.

1. Gather data (run all independent steps in parallel)

  • git log --format="%h %s (%ar)" -20, git status, and git worktree list
  • Read .branch-context.md (if exists — check CWD and worktree root). Parse YAML frontmatter: workflow_step, validate_round/validate_max, review_round/review_max, plan_review_chunk/plan_review_total/plan_chunks_approved
  • Read ai-workspace/MEMORY.md — note anything in-progress or blocking
  • Read HANDOFF.md (if exists)
  • List ai-workspace/plans/ (sort by mtime, most recent first) and read the first file that does NOT end in .done.md or .executed.md
  • Search Basic Memory via MCP for recent entries matching the current branch name. If MCP unavailable, skip silently.
  • GitHub — check availability first:
command -v gh >/dev/null 2>&1 && gh auth status >/dev/null 2>&1

If authenticated, run in parallel:

gh pr list --state open

gh issue list --state open --limit 15

If gh is unavailable or unauthenticated, note "GitHub CLI unavailable" and skip. Do NOT attempt to install or authenticate.

2. Synthesize

Present findings in this format. Omit sections where the data source was not found (e.g., no HANDOFF.md, gh unavailable). Always show Loop state and Blockers even when empty — absence is informative:

Branch:         <name> — <what we're building (from plan or branch-context)>

Worktree:       <path> (or "primary")

Last commit:    <hash> <message>

Uncommitted:    <staged/unstaged summary, or "clean">

Workflow step:  <N> — <step name> (from frontmatter, or inferred)

Loop state:     validate <round>/<max>, review <round>/<max>

Active plan:    <file> — <current phase> — <next action>

Blockers:       <list>

Open PRs:       <count — titles>

Open issues:    <count — top 5 titles>

Handoff:        <detail>

After the summary, surface blockers and TODO items prominently — remaining work, failing CI, stale branches, items from memory flagged as blocking. These go above the fold, not as footnotes.

3. Suggest next actions

Based on gathered state, suggest 2-3 concrete next actions. Examples:

  • "Resume building — plan X is at step 5 (Build)"
  • "Run /validate — implementation looks complete"
  • "Address blocker: [description]"
  • "Review open PR #N — awaiting action"
  • "Pick up issue #N — in-progress"

End with: "What would you like to do?"

Failure modes

  • **No .branch-context.md**: infer workflow step from git state and plan file status. Note: "No branch context — step inferred."
  • No active plan: note "No active plan." Do not search .done.md files.
  • **On main branch**: skip branch-context and plan lookup. Show git state, PRs, issues, and memory only.
  • Basic Memory MCP unavailable: skip silently.
  • **gh unavailable**: show "GitHub CLI unavailable — PR/issue status unknown."
BrowserAct

Let your agent run on any real-world website

Bypass CAPTCHA & anti-bot for free. Start local, scale to cloud.

Explore BrowserAct Skills →

Stop writing automation&scrapers

Install the CLI. Run your first Skill in 30 seconds. Scale when you're ready.

Start free
free · no credit card