add Context and Session Management tips sections from Thariq 16 April article
- split 10 context-management tips into Context (5) + Session Management (5) - move /compact, /rewind, /rename tips out of Workflows into the new sections - remove bold and inline-code formatting from tip content per style cleanup - update TOC and section counts (total tips: 69 → 76) Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Shayan Rais
@@ -122,11 +122,11 @@ All major workflows converge on the same architectural pattern: **Research → P
|
||||
<img src="!/claude-jumping.svg" alt="section divider" width="60" height="50">
|
||||
</p>
|
||||
|
||||
## 💡 TIPS AND TRICKS (69)
|
||||
## 💡 TIPS AND TRICKS (76)
|
||||
|
||||
🚫👶 = do not babysit
|
||||
|
||||
[Prompting](#tips-prompting) · [Planning](#tips-planning) · [CLAUDE.md](#tips-claudemd) · [Agents](#tips-agents) · [Commands](#tips-commands) · [Skills](#tips-skills) · [Hooks](#tips-hooks) · [Workflows](#tips-workflows) · [Advanced](#tips-workflows-advanced) · [Git / PR](#tips-git-pr) · [Debugging](#tips-debugging) · [Utilities](#tips-utilities) · [Daily](#tips-daily)
|
||||
[Prompting](#tips-prompting) · [Planning](#tips-planning) · [Context](#tips-context) · [Session](#tips-session) · [CLAUDE.md](#tips-claudemd) · [Agents](#tips-agents) · [Commands](#tips-commands) · [Skills](#tips-skills) · [Hooks](#tips-hooks) · [Workflows](#tips-workflows) · [Advanced](#tips-workflows-advanced) · [Git / PR](#tips-git-pr) · [Debugging](#tips-debugging) · [Utilities](#tips-utilities) · [Daily](#tips-daily)
|
||||
|
||||
|
||||
|
||||
@@ -149,6 +149,26 @@ All major workflows converge on the same architectural pattern: **Research → P
|
||||
| write detailed specs and reduce ambiguity before handing work off — the more specific you are, the better the output | [](https://x.com/bcherny/status/2017742752566632544) |
|
||||
| prototype > PRD — build 20-30 versions instead of writing specs, the cost of building is low so take many shots | [](https://youtu.be/julbw1JuAz0?t=3630) [](https://youtu.be/julbw1JuAz0?t=3630) |
|
||||
|
||||
<a id="tips-context"></a>■ **Context (5)**
|
||||
|
||||
| Tip | Source |
|
||||
|-----|--------|
|
||||
| context rot kicks in around ~300-400k tokens on the 1M context model — don't let sessions drift past that for intelligence-sensitive work | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| avoid agent dumb zone, do manual [/compact](https://code.claude.com/docs/en/interactive-mode) at max 50%. Use [/clear](https://code.claude.com/docs/en/cli-reference) to reset context mid-session if switching to a new task | |
|
||||
| rewind > correct — double-Esc or [/rewind](https://code.claude.com/docs/en/checkpointing) back to before the failed attempt and re-prompt with what you learned, instead of leaving failed attempts + corrections polluting context 🚫👶 | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| [/compact](https://code.claude.com/docs/en/interactive-mode) with a hint (/compact focus on the auth refactor, drop the test debugging) beats letting autocompact fire — the model is at its least intelligent point when auto-compacting due to context rot | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| use subagents for context management — ask yourself "will I need this tool output again, or just the conclusion?" — 20 file reads + 12 greps + 3 dead ends stay in the child's context, only the final report returns 🚫👶 | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
|
||||
<a id="tips-session"></a>■ **Session Management (5)**
|
||||
|
||||
| Tip | Source |
|
||||
|-----|--------|
|
||||
| every turn is a branching point — after Claude ends a turn, pick between Continue, /rewind, /clear, /compact, or Subagent based on how much existing context you need to carry forward | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| new task = new session — related tasks (e.g. writing docs for what you just built) can reuse context for efficiency, but genuinely new tasks deserve a fresh session | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| use "summarize from here" before rewinding to have Claude write a handoff message — like a note to the previous iteration of Claude from its future self | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| /compact vs /clear — compact is lossy but momentum-friendly (mid-task, fuzzy details ok); /clear + brief is more work but you control exactly what carries forward (high-stakes next step) | [](tips/claude-thariq-tips-16-apr-26.md) |
|
||||
| [/rename](https://code.claude.com/docs/en/cli-reference) important sessions (e.g. [TODO - refactor task]) and [/resume](https://code.claude.com/docs/en/cli-reference) them later — label each instance when running multiple Claudes simultaneously | [](https://every.to/podcast/how-to-use-claude-code-like-the-people-who-built-it) |
|
||||
|
||||
<a id="tips-claudemd"></a>■ **CLAUDE.md (7)**
|
||||
|
||||
| Tip | Source |
|
||||
@@ -160,7 +180,7 @@ All major workflows converge on the same architectural pattern: **Research → P
|
||||
| [memory.md](https://code.claude.com/docs/en/memory), constitution.md does not guarantee anything | |
|
||||
| any developer should be able to launch Claude, say "run the tests" and it works on the first try — if it doesn't, your CLAUDE.md is missing essential setup/build/test commands | [](https://x.com/dexhorthy/status/2034713765401551053) |
|
||||
| keep codebases clean and finish migrations — partially migrated frameworks confuse models that might pick the wrong pattern | [](https://youtu.be/julbw1JuAz0?t=1112) [](https://youtu.be/julbw1JuAz0?t=1112) |
|
||||
| use [settings.json](best-practice/claude-settings.md) for harness-enforced behavior (attribution, permissions, model) — don't put "NEVER add Co-Authored-By" in CLAUDE.md when `attribution.commit: ""` is deterministic | [](https://x.com/dani_avila7/status/2036182734310195550) |
|
||||
| use [settings.json](best-practice/claude-settings.md) for harness-enforced behavior (attribution, permissions, model) — don't put "NEVER add Co-Authored-By" in CLAUDE.md when attribution.commit: "" is deterministic | [](https://x.com/dani_avila7/status/2036182734310195550) |
|
||||
|
||||
<a id="tips-agents"></a><img src="!/tags/a.svg" height="14"> **Agents (4)**
|
||||
|
||||
@@ -176,8 +196,8 @@ All major workflows converge on the same architectural pattern: **Research → P
|
||||
| Tip | Source |
|
||||
|-----|--------|
|
||||
| use [commands](https://code.claude.com/docs/en/slash-commands) for your workflows instead of [sub-agents](https://code.claude.com/docs/en/sub-agents) | [](https://x.com/bcherny/status/2007179847949500714) |
|
||||
| use [slash commands](https://code.claude.com/docs/en/slash-commands) for every "inner loop" workflow you do many times a day — saves repeated prompting, commands live in `.claude/commands/` and are checked into git | [](https://x.com/bcherny/status/2007179847949500714) |
|
||||
| if you do something more than once a day, turn it into a [skill](https://code.claude.com/docs/en/skills) or [command](https://code.claude.com/docs/en/slash-commands) — build `/techdebt`, context-dump, or analytics commands | [](https://x.com/bcherny/status/2017742748984742078) |
|
||||
| use [slash commands](https://code.claude.com/docs/en/slash-commands) for every "inner loop" workflow you do many times a day — saves repeated prompting, commands live in .claude/commands/ and are checked into git | [](https://x.com/bcherny/status/2007179847949500714) |
|
||||
| if you do something more than once a day, turn it into a [skill](https://code.claude.com/docs/en/skills) or [command](https://code.claude.com/docs/en/slash-commands) — build /techdebt, context-dump, or analytics commands | [](https://x.com/bcherny/status/2017742748984742078) |
|
||||
|
||||
<a id="tips-skills"></a><img src="!/tags/s.svg" height="14"> **Skills (9)**
|
||||
|
||||
@@ -191,7 +211,7 @@ All major workflows converge on the same architectural pattern: **Research → P
|
||||
| don't state the obvious in skills — focus on what pushes Claude out of its default behavior 🚫👶 | [](https://x.com/trq212/status/2033949937936085378) |
|
||||
| don't railroad Claude in skills — give goals and constraints, not prescriptive step-by-step instructions 🚫👶 | [](https://x.com/trq212/status/2033949937936085378) |
|
||||
| include scripts and libraries in skills so Claude composes rather than reconstructs boilerplate | [](https://x.com/trq212/status/2033949937936085378) |
|
||||
| embed `` !`command` `` in SKILL.md to inject dynamic shell output into the prompt — Claude runs it on invocation and the model only sees the result | [](https://x.com/lydiahallie/status/2034337963820327017) |
|
||||
| embed !command in SKILL.md to inject dynamic shell output into the prompt — Claude runs it on invocation and the model only sees the result | [](https://x.com/lydiahallie/status/2034337963820327017) |
|
||||
|
||||
<a id="tips-hooks"></a>■ **Hooks (5)**
|
||||
|
||||
@@ -203,17 +223,14 @@ All major workflows converge on the same architectural pattern: **Research → P
|
||||
| route [permission requests](https://code.claude.com/docs/en/hooks) to Opus via a hook — let it scan for attacks and auto-approve safe ones 🚫👶 | [](https://x.com/bcherny/status/2017742755737555434) |
|
||||
| use a [Stop hook](https://code.claude.com/docs/en/hooks) to nudge Claude to keep going or verify its work at the end of a turn | [](https://x.com/bcherny/status/2021701059253874861) |
|
||||
|
||||
<a id="tips-workflows"></a>■ **Workflows (7)**
|
||||
<a id="tips-workflows"></a>■ **Workflows (4)**
|
||||
|
||||
| Tip | Source |
|
||||
|-----|--------|
|
||||
| avoid agent dumb zone, do manual [/compact](https://code.claude.com/docs/en/interactive-mode) at max 50%. Use [/clear](https://code.claude.com/docs/en/cli-reference) to reset context mid-session if switching to a new task | |
|
||||
| vanilla cc is better than any workflows with smaller tasks | |
|
||||
| use [/model](https://code.claude.com/docs/en/model-config) to select model and reasoning, [/context](https://code.claude.com/docs/en/interactive-mode) to see context usage, [/usage](https://code.claude.com/docs/en/costs) to check plan limits, [/extra-usage](https://code.claude.com/docs/en/interactive-mode) to configure overflow billing, [/config](https://code.claude.com/docs/en/settings) to configure settings — use Opus for plan mode and Sonnet for code to get the best of both | [](https://x.com/_catwu/status/1955694117264261609) |
|
||||
| always use [thinking mode](https://code.claude.com/docs/en/model-config) true (to see reasoning) and [Output Style](https://code.claude.com/docs/en/output-styles) Explanatory (to see detailed output with ★ Insight boxes) in [/config](https://code.claude.com/docs/en/settings) for better understanding of Claude's decisions | [](https://x.com/bcherny/status/2007179838864666847) |
|
||||
| use ultrathink keyword in prompts for [high effort reasoning](https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#tips-and-best-practices) | |
|
||||
| [/rename](https://code.claude.com/docs/en/cli-reference) important sessions (e.g. [TODO - refactor task]) and [/resume](https://code.claude.com/docs/en/cli-reference) them later — label each instance when running multiple Claudes simultaneously | [](https://every.to/podcast/how-to-use-claude-code-like-the-people-who-built-it) |
|
||||
| use [Esc Esc or /rewind](https://code.claude.com/docs/en/checkpointing) to undo when Claude goes off-track instead of trying to fix it in the same context | |
|
||||
|
||||
<a id="tips-workflows-advanced"></a>■ **Workflows Advanced (6)**
|
||||
|
||||
|
||||
Reference in New Issue
Block a user