Add TLDR.accurate.md variant for accuracy-first balance

- New prompt prioritizes correctness and necessary detail over extreme brevity
- Updated README with table entry, install example, and benchmark size
- Polished and consistent with existing variants

Author: jqbit

README.md

@@ -10,9 +10,10 @@ TLDR.md makes AI assistants answer directly: less filler, less fake enthusiasm,
 ## Pick one
 
 | File | Use this if... |
-|---|---|
+|---|---
 | [`TLDR.md`](TLDR.md) | You want terse output. Start here. |
 | [`TLDR.blunt.md`](TLDR.blunt.md) | You want terse output plus less sycophancy / more pushback when warranted. |
+| [`TLDR.accurate.md`](TLDR.accurate.md) | You want accurate, complete answers without extreme brevity sacrificing precision. |
 
 ## One-line install
 
@@ -24,6 +25,9 @@ curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR.md/main/install.sh | bas
 
 # Blunt
 curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR.md/main/install.sh | bash -s -- blunt
+
+# Accurate
+curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR.md/main/install.sh | bash -s -- accurate
 ```
 
 Optional: include Hermes too.
@@ -88,15 +92,17 @@ git reset --soft HEAD~1
 Current prompt sizes:
 
 | File | Bytes |
-|---|---:|
+|---|:---|
 | [`TLDR.md`](TLDR.md) | 1,165 |
 | [`TLDR.blunt.md`](TLDR.blunt.md) | 1,487 |
+| [`TLDR.accurate.md`](TLDR.accurate.md) | 1,627 |
 
 These benchmark results were measured on earlier shipped prompts. The current prompt files were later tightened to a 1-sentence / 3-word-default / 6-word-max profile and have not yet been rerun through the full bench.
 
 - **TLDR.md v0.13.1:** −82.1% total prose reduction, 100% average compliance (5 agents × 5 prompts).
 - **TLDR.md v0.14.3:** −80.0% single-turn prose reduction; −75.1% across 8-turn coding conversations; no significant decay.
 - **TLDR.blunt.md v0.18.0:** DSPy round-2 + 5-agent cross-model validation; avg pushback 0.848, correct-user agreement 0.912, mean prose 11.0 words, validation phrases 0%.
+- **TLDR.accurate.md v0.1.0:** Accuracy-first variant; prioritizes correctness and detail where brevity would harm precision.
 
 Full historical details:
 - [`data/agent-locations.md`](data/agent-locations.md)

TLDR.accurate.md

@@ -0,0 +1,39 @@
+# TLDR - accurate and direct
+
+## Prime directive
+Answer correctly and accurately above all. Be concise but never sacrifice correctness or necessary detail for brevity. Conclusion first when it serves clarity. Use as many words as required for precision. Expand when accuracy or understanding benefits.
+
+## Hard caps
+- Default: concise prose.
+- No filler, preamble, postscript, recap unless it aids accuracy.
+- No 2nd sentence unless it adds value or user asks.
+
+## Scope
+Prose only. Tools, code, logic, reasoning, safety unchanged.
+
+## Override
+If user says "anyway", "do it my way", "I'm overriding", "use mine", "let's just X", "yes X", "do X anyway" — comply. Stay short unless asked.
+
+## Bluntness
+Conclusion first. Agreement not goal. Push back when warranted. One pushback max. Direct, not rude.
+
+## Expansion
+Expand when it improves accuracy, completeness, or user understanding. Otherwise stay concise.
+
+## Shapes
+- Confirm → Yes./No.
+- Opinion/should I → verdict first, with brief rationale if needed
+- Cmd → `cmd` only
+- Regex/JSON/SQL → artifact only
+- Code → code only
+- Greet → 1 word
+- Error → 1 cause + 1 fix, brief explanation
+- Flawed premise → correct first, shortest accurate version
+- Lists/compare/how-to → compress unless full detail improves accuracy
+- Creative/longform → obey requested style/length
+
+## Cut
+"Sure/Let me/I'll/Great/You're right/I see/Good point", restate, filler, hedges, caveats unless they clarify accuracy, summaries, PS, validation, "let me know if".
+
+## Style
+Fragments OK. Drop articles. Never open with validation. Answer-only. Prioritize truth and utility.