fix(keybindings): correct Ctrl+R from "Retry" to "Search history"

- Updated 5 locations in guide (cheatsheet + ultimate-guide)
- Verified against official keybindings: history:search action
- Added resource evaluation: Sankalp's Claude Code experience (2/5)
- Blog correctly identified guide error

Closes evaluation workflow for sankalp-claude-code-experience

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit is contained in:
Florian BRUNIAUX 2026-02-03 17:27:54 +01:00
parent 65ce82758f
commit 0a2e05f290
4 changed files with 179 additions and 6 deletions

View file

@ -58,7 +58,8 @@ Les documents de travail bruts (prompts Perplexity, audits clients) restent dans
| **Docker Sandboxes** (Isolation Landscape) | 4/5 | **4/5** | ✅ Intégré (guide + notice) | [docker-sandboxes-isolation.md](./docker-sandboxes-isolation.md) |
| **dclaude** (Dockerized Claude Code) | 2/5 | **2/5** | ⚠️ Footnote (sandbox-isolation.md) | [dclaude-docker-wrapper.md](./dclaude-docker-wrapper.md) |
| **10 Tips from Inside the Claude Code Team** (paddo.dev) | 4/5 | **4/5** | ✅ Intégré (4 sections) | [paddo-team-tips-eval.md](./paddo-team-tips-eval.md) |
| **Sankalp's Claude Code 2.0 Experience** | 2/5 | **2/5** | ⚠️ Watch only (85% overlap, probable errors) | [sankalp-claude-code-experience.md](./sankalp-claude-code-experience.md) |
---
**Dernier update**: 2026-02-01 (20 évaluations)
**Dernier update**: 2026-02-03 (21 évaluations)

View file

@ -0,0 +1,172 @@
# Resource Evaluation: Sankalp's "My experience with Claude Code 2.0"
**URL**: https://sankalp.bearblog.dev/my-experience-with-claude-code-20-and-how-to-get-better-at-using-coding-agents/
**Author**: Sankalp (@dejavucoder)
**Date**: December 27, 2025
**Type**: Technical deep-dive blog post with experience narrative
**Evaluated**: 2026-02-03
**Score**: 2/5 (Marginal)
---
## Executive Summary
Experience-based blog covering Claude Code 2.0 features with narrative framing. **85% overlap** with existing guide content, but guide is more precise on every shared topic. Two potentially new items fail scrutiny: 50-60% context claim conflates distinct concepts, and model comparisons are deliberately out of scope. **One correction found**: Blog correctly identifies Ctrl+R as "history search" — our guide had documented it incorrectly as "Retry" (now fixed).
**Decision**: Watch-only. No integration.
---
## Content Summary
Technical blog covering:
- Context engineering principles and effective capacity
- Sub-agents, checkpointing, plugins, hooks, skills
- Model comparisons (Opus 4.5 vs GPT-5.2-Codex vs Gemini 3 Pro)
- Claims effective context windows operate at ~50-60% of stated capacity
- Claims Ctrl+R is "history search" (similar to terminal backsearch)
- References Anthropic's "Building Effective Agents", Chroma context rot research, Manus.im
- No specific SWE-bench scores despite mentioning benchmark
- No stated credentials beyond Claude Code user since June 2025
---
## Gap Analysis
| Topic | Guide Coverage | Blog Adds? |
|-------|---------------|------------|
| Context engineering | **EXTENSIVE** (architecture.md:259-298, ultimate-guide.md:1550-1681) | Different framing, less precise |
| Sub-agents | **COMPREHENSIVE** (architecture.md:395-447) | Nothing — guide more detailed |
| Checkpointing/rewind | **COMPREHENSIVE** (ultimate-guide.md:2405-2535) | Nothing |
| Plugins system | **COMPREHENSIVE** (ultimate-guide.md:9177-9430) | Nothing |
| Tool call overhead | **COMPREHENSIVE** (architecture.md:166-175, ultimate-guide.md:1661-1669) | Nothing |
| Hooks lifecycle | **COMPREHENSIVE** (775+ lines, 25+ templates) | Nothing |
| Skills loading | **COVERED** (ultimate-guide.md:5440-5498) | Narrative framing only |
| Model comparisons | **PARTIAL** (deliberate exclusion) | Yes, but out of scope |
| Context window sizes | **PARTIAL** (200K only) | 400K/1M data, but stale quickly |
| Ctrl+R / prompt history | **Guide said "Retry"** (5 locations, now corrected) | ✅ **Blog correct** — identified guide error |
**Overlap**: ~85% of content already in guide, and guide is more precise on every shared topic.
---
## Fact-Check Results
| Claim | Status | Verification |
|-------|--------|--------------|
| Author: Sankalp (@dejavucoder) | ✅ Verified | Article byline |
| Date: Dec 27, 2025 | ✅ Verified | Article header |
| Opus 4.5 has 200K context window | ✅ Verified | Consistent with guide (architecture.md:37) |
| GPT-5.2 has 400K input tokens | ⚠️ Unverified | Plausible but not independently confirmed |
| Gemini 3 Pro has 1M context | ⚠️ Unverified | Plausible but specific version needs verification |
| Effective context at 50-60% | ❌ **Incorrect** | Conflates usable capacity (70-75% per architecture.md:295) with quality degradation threshold. Guide's Chroma Research reference shows 20-30% performance gap but at higher fill rates |
| ~50 tool calls average (Manus) | ⚠️ Unverified | Attributed to Manus.im blog, not independently verified |
| Ctrl+R = history search | ✅ **Verified - Blog correct, guide error** | Official keybindings confirm `history:search` action. Our guide incorrectly documented as "Retry" in 5 locations (now corrected) |
| Syntax highlighting in 2.0.71 | ⚠️ Unclear | Version numbering format doesn't match our releases tracking |
| Opus 4.5 SOTA on SWE-bench | ⚠️ Incomplete | Mentioned but no specific scores provided |
| GPT-5.2-Codex exceeds Opus 4.5 | ⚠️ Opinion | Opinion-level claim, no benchmark data cited |
### Critical Issues
**1. Context Capacity Claim**: Blog conflates two distinct concepts:
- **Usable capacity** (70-75%): Maximum recommended fill rate for performance
- **Quality degradation**: Performance gap that appears at high fill rates
The guide handles both concepts separately and correctly. The 50-60% figure oversimplifies and risks misguiding users.
### Correction Applied to Guide
**Ctrl+R Keybinding**: Blog correctly identifies Ctrl+R as "history search". Our guide had incorrectly documented it as "Retry" in 5 locations. Verification against official keybindings (`history:search` action) confirms the blog was right. Guide has been corrected (cheatsheet.md + ultimate-guide.md).
---
## Technical Writer Challenge Summary
The technical-writer agent evaluated the blog and confirmed the 2/5 score, noting it's "possibly generous":
### Key Findings
- **Ctrl+R claim**: ✅ **Blog was correct** — identified an error in our guide (now fixed)
- **50-60% effective capacity**: Blog conflates two distinct concepts the guide handles separately
- **Narrative format**: Does not clear the Practitioner Insights bar:
- No production-scale validation
- No novel patterns demonstrated
- No credentials stated beyond "Claude Code user since June 2025"
- **Temporal decay risk**: Model comparisons will be stale quickly
- **Source credibility**: Blog lacks authoritative backing
### Verdict
Watch-only. Minimal integration (Ctrl+R correction applied). The blog provides a user narrative and identified one guide error, but adds no other verified information beyond what's already covered.
---
## Scoring Justification
**Score: 2/5 (Marginal)**
| Criterion | Assessment |
|-----------|------------|
| **Accuracy** | Mixed — 1 error (context capacity), 1 correction provided (Ctrl+R), several unverified claims |
| **Novelty** | Low — 85% overlap with existing guide content |
| **Depth** | Moderate — Narrative format, less precise than guide |
| **Credibility** | Low — No credentials, several unverified claims |
| **Actionability** | Low — No novel patterns or techniques |
### Why Not Lower?
The blog isn't fundamentally misleading — most content aligns with known Claude Code features. It simply doesn't add value beyond what the guide already covers more thoroughly.
### Why Not Higher?
- 7/10 topics already covered comprehensively
- 2/3 potentially new items fail scrutiny, 1 correct (Ctrl+R) but minor impact
- No production-scale validation
- Temporal decay risk for model comparisons
- Source credibility concerns
---
## Decision
**Action**: Watch-only. Minimal integration (Ctrl+R correction applied).
**Rationale**:
1. One correction applied (Ctrl+R keybinding now accurate in guide)
2. All other topics already covered more thoroughly in guide
3. One error remaining (context capacity oversimplification)
4. No credentials or production-scale validation
**Confidence**: High (fact-check confirms no actionable gaps)
---
## Follow-up Completed
### CLI Test Results
**Action completed**: Tested Ctrl+R in Claude Code CLI and verified against official keybindings.
**Result**: ✅ Blog was **correct** — Ctrl+R triggers `history:search` action (Global context) and `historySearch:next` (HistorySearch context).
**Guide corrections applied**:
- `guide/cheatsheet.md:39` — Updated from "Retry last operation" to "Search command history"
- `guide/ultimate-guide.md:358` — Updated from "Retry last operation" to "Search command history"
- `guide/ultimate-guide.md:15508` — Updated from "Retry last" to "Search history"
- `guide/ultimate-guide.md:15521` — Updated from "Retry last operation" to "Search command history"
- `guide/ultimate-guide.md:16032` — Updated from "Retry" to "Search" in ASCII art box
---
## Archive Notes
- **Watch status**: Monitor for future updates or corrections from author
- **Temporal sensitivity**: Model comparison data will decay quickly
- **Community value**: May be useful as user narrative, not as technical reference
- **Guide impact**: Minor — Ctrl+R keybinding corrected in 5 locations (cheatsheet + ultimate-guide)
---
**Evaluation completed**: 2026-02-03
**Corrections applied**: 2026-02-03 (Ctrl+R keybinding)
**Next review**: Not scheduled (watch-only status)

View file

@ -36,7 +36,7 @@
| `Shift+Tab` | Cycle permission modes |
| `Esc` × 2 | Rewind (undo) |
| `Ctrl+C` | Interrupt |
| `Ctrl+R` | Retry last operation |
| `Ctrl+R` | Search command history |
| `Ctrl+L` | Clear screen (keeps context) |
| `Tab` | Autocomplete |
| `Shift+Enter` | New line |

View file

@ -355,7 +355,7 @@ These 7 commands are the ones I use most frequently:
| `!command` | Run shell command directly | `!git status`, `!npm test` |
| `@file.ts` | Reference a specific file | `@src/app.tsx`, `@README.md` |
| `Ctrl+C` | Cancel current operation | Stop long-running analysis |
| `Ctrl+R` | Retry last operation | Retry after fixing error |
| `Ctrl+R` | Search command history | Find previous prompts |
| `Esc` | Dismiss current suggestion | Skip unwanted changes |
#### Shell Commands with `!`
@ -15505,7 +15505,7 @@ _Quick jump:_ [Commands Table](#101-commands-table) · [Keyboard Shortcuts](#102
| Run shell command | `!command` |
| Reference file | `@filename` |
| Cancel operation | `Ctrl+C` |
| Retry last | `Ctrl+R` |
| Search history | `Ctrl+R` |
| Dismiss suggestion | `Esc` |
## 10.2 Keyboard Shortcuts
@ -15518,7 +15518,7 @@ _Quick jump:_ [Commands Table](#101-commands-table) · [Keyboard Shortcuts](#102
| `Shift+Enter` | New line in message |
| `Ctrl+C` | Cancel current operation |
| `Ctrl+D` | Exit Claude Code |
| `Ctrl+R` | Retry last operation |
| `Ctrl+R` | Search command history |
| `Ctrl+L` | Clear screen (keeps context) |
| `Ctrl+B` | Run command in background |
| `Esc` | Dismiss current suggestion |
@ -16029,7 +16029,7 @@ Get the scripts from:
║ QUICK ACTIONS ║
║ ───────────── ║
║ !command Run shell @file Reference file ║
║ Ctrl+R Retry ↑/↓ History ║
║ Ctrl+R Search ↑/↓ History ║
║ ║
║ CONTEXT MANAGEMENT ║
║ ────────────────── ║