Terse, high-signal responses for AI coding agents — less filler, same accuracy on tools, code, and safety.
One repo, two install paths:
| Path | What you get | Best for |
|---|
TLDR.md + install.sh | Copies the system prompt (and /tldr command) into standard agent config locations | Quick global rules file, no Node required |
bin/install.js (via npx) | Detects installed agents; plugins, hooks, skills | Claude Code, Cursor, Codex, Gemini, 30+ agents — see docs/INSTALL.md |
Historical prompt variants and benchmarks: data/changelog.md, data/progression.md.
Before / After
🗣️ Normal agent (verbose)
"Sure! I'd be happy to help. The reason your React component is re-rendering is likely because you're creating a new object reference on each render. I'd recommend wrapping it in useMemo."
TLDR mode
New object ref each render. Inline prop = new ref = re-render. useMemo.
Same fix. ~60–75% fewer prose tokens in historical benchmarks (earlier prompt generations; the current prompt has not been rerun) — tools, code, and safety unchanged. Writeup: data/benchmarks.md · honest cost/benefit: docs/HONEST-NUMBERS.md.
The prompt — TLDR.md
TLDR.md is the active prompt (2,218 bytes). It changes prose style only — not tools, reasoning, or safety.
Current defaults (prompt):
- default: 1 sentence
- target: 3 words
- default max: 6 words
- one-word greeting for plain greetings
/tldr (supported agents) re-applies rules live in long sessions- TLDR footer: replies to the user only that exceed 3 sentences, 30 words, or 100 characters end with a brief
TLDR: summary line
Install — prompt only (install.sh)
No Node required. Writes TLDR.md to seven standard agent paths and installs /tldr where supported.
curl -fsSL https://raw.githubusercontent.com/0p9b/TLDR/main/install.sh | bash -s --
Optional Hermes merge into ~/.hermes/SOUL.md:
curl -fsSL https://raw.githubusercontent.com/0p9b/TLDR/main/install.sh | bash -s -- --with-hermes
(Prompt-only. Full Hermes skill suite → node bin/install.js --only hermes.)
Preview: curl -fsSL https://raw.githubusercontent.com/0p9b/TLDR/main/install.sh
Manual paths and copy/paste commands: data/agent-locations.md.
Install — full stack (bin/install.js)
Node ≥18. Auto-detects agents, installs the right plugin/extension/skills/hooks per agent.
npx -y github:0p9b/TLDR
From a clone:
git clone https://github.com/0p9b/TLDR.git && cd TLDR
node bin/install.js # detected agents
node bin/install.js --all # hooks
node bin/install.js --list # agent matrix
Windows: install.ps1 runs the Node installer.
Full flags, per-agent table, verify, uninstall, troubleshooting: docs/INSTALL.md.
Updating
tldr update # or: node bin/install.js update
tldr update --check # report available update only
Refreshes the git tree (local clone or ~/.tldr/src), then reinstalls agent hooks/skills. Details: docs/INSTALL.md § Updating.
Verify (prompt install)
for p in ~/.claude/CLAUDE.md ~/.gemini/AGENTS.md ~/.codex/AGENTS.md \
~/AGENTS.md ~/.config/opencode/AGENTS.md \
~/.factory/AGENTS.md ~/.pi/agent/AGENTS.md; do
[ -f "$p" ] && grep -q "^## Prime directive" "$p" && echo "✓ $p" || echo "✗ $p"
done
grep -q "^## Prime directive" ~/.hermes/SOUL.md 2>/dev/null && echo "✓ ~/.hermes/SOUL.md" || echo "✗ ~/.hermes/SOUL.md"
Repository map
| Path | Purpose |
|---|
TLDR.md | Canonical terse system prompt |
commands/tldr.md | /tldr slash command |
install.sh | Prompt-only installer (+ optional Hermes) |
install.ps1 | Windows entry → bin/install.js |
bin/install.js | Unified multi-agent installer |
skills/ | TLDR skill suite (source of truth for behavior) |
plugins/tldr/ | Claude Code / distribution mirrors (CI-synced) |
src/hooks/ | Claude Code SessionStart / mode hooks |
docs/INSTALL.md | Full installer documentation |
.github/CONTRIBUTING.md | How to change skills or add agents |
data/agent-locations.md | Per-agent install paths |
Research & benchmarks
License
MIT — see LICENSE. Third-party lineage/notices: ATTRIBUTION.md, THIRD_PARTY_NOTICES.md.
