agentsclimarketplace

Catchup

Skill SamyakJhaveri/loam/seed/.claude/skills/catchup

Copier template that bootstraps AI-agent-optimized project setups: layered context routing, curated skills, and an enforced validation gate — synced across projects.

Install
npx -y skills add SamyakJhaveri/loam --skill catchup

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

  • 0 stars0 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.

What its author says it does

Copied from the file, not written here

Fast 30s session bootstrap briefing. Use when resuming work after any break, at the start of a fresh session, or when unsure of current project state. Reports git status, recent commits, env state, memory staleness, pending tasks, and red flags (uncommitted changes, detached HEAD, stale memory). NOT for: deep codebase exploration, implementing changes, or running tests — only a read-only status snapshot.

SKILL.md

4.9 KB, as published. Nobody here has run it

Session Catchup Briefing

Use when resuming work after a break and you need a fast 30-second context refresh. Lighter than /session-start — no prerequisite checks, no sprint context, just the state of the repo and any in-flight work.

Trigger: When user types /catchup.

Iron Law

NO WORK WITHOUT CONTEXT — READ THE STATE BEFORE CHANGING IT

Arguments

  • $ARGUMENTS — optional: number of recent commits to show (default: 10)

Anti-Rationalization Table

ExcuseReality
"I remember what I was doing"Memory decays; git log doesn't. 30 seconds now saves 30 minutes of confusion later
"I'll figure it out as I go"Figuring it out mid-task means wasted context window on re-discovery
"Nothing changed since last session"Verify that claim — auto-merges, hook updates, and eval batches run silently
"I just have a quick fix"Quick fixes on stale context cause the bugs that take hours to debug

Red Flags — STOP and Warn User

  • Uncommitted changes in the working tree (risk of losing work or committing stale state)
  • Detached HEAD state (not on a named branch)
  • Memory files older than 14 days with no updates (context may be stale)
  • Active tmux sessions that may be running eval batches (avoid conflicts)
  • Merge conflicts or rebase in progress

If any red flag triggers: display it prominently at the TOP of the briefing, before the normal status report.

Project Context

  • Project root: auto-detect via git rev-parse --show-toplevel
  • Venv: auto-detect (.venv/, venv/, env/)
  • Memory directory: auto-detect from ~/.claude/projects/ (based on project root path)
  • Check CLAUDE.md for project-specific context (key counts, models, etc.)

Workflow

Phase 1: Git State

Run these commands and capture output — do NOT guess or use cached data:

# Current branch and dirty state
cd "$(git rev-parse --show-toplevel)" && git status

# Recent commits (default 10, or user-specified count)
git log --oneline -<N>

# Changed files since ~last session (diff stat against 10 commits back)
git diff --stat HEAD~10 2>/dev/null || git diff --stat $(git rev-list --max-parents=0 HEAD)

Verification gate: All three commands must execute successfully. If git status shows an error (not a repo, corrupted index), STOP and report.

Phase 2: Environment Check

# Check for running tmux sessions (eval batches, overnight runs)
tmux list-sessions 2>/dev/null || echo "No tmux sessions"

# Check if venv is active or needs activation
python3 --version 2>/dev/null
which python3 2>/dev/null

Verification gate: If tmux shows active sessions, flag them — the user may have an eval batch running that should not be interrupted.

Phase 3: Memory Staleness

Read the memory index at the memory directory path and check dates:

# Auto-detect memory directory from project root
PROJECT_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || pwd)
ENCODED_PATH=$(echo "$PROJECT_ROOT" | sed 's|/|-|g')
MEMORY_DIR="$HOME/.claude/projects/$ENCODED_PATH/memory"
ls -lt "$MEMORY_DIR"/*.md 2>/dev/null | head -10

Flag any memory file not modified in 14+ days as potentially stale.

Verification gate: Memory index file must exist. If not, note that memory system needs initialization.

Phase 4: Task State

Run TaskList to check for any in-progress or pending tasks from a previous session.

Verification gate: If tasks exist from a previous session, display them — they may represent interrupted work.

Phase 5: Compile Briefing

Present a concise bullet-point briefing in this exact format:

=== CATCHUP BRIEFING ===

[RED FLAGS — only if any detected]
  ! Uncommitted changes: <N> files modified
  ! Active tmux session: <session name> (may be running eval batch)

Branch:    <current branch>
Status:    clean | <N> modified, <N> untracked
Last commit: <hash> <message> (<relative time>)

Recent activity (last <N> commits):
  <hash> <message>
  <hash> <message>
  ...

Changed files (since HEAD~10):
  <diffstat summary>

Environment:
  tmux:   <N sessions active | none>
  venv:   <active | needs activation>
  python: <version>

Memory:
  <N> files, most recent: <filename> (<date>)
  [Stale: <list of files >14 days old>]

Open tasks:
  [#<id>] <status> — <subject>
  ...
  [or: No open tasks]

=== END BRIEFING ===

Keep it tight. The entire briefing should fit in one screen. No explanations, no suggestions, no preamble — just the state. The user decides what to do with it.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.