Prompting Claude 5: What Changed, What Didn't, and What People Get Wrong
1. Task
Start with why, NOT what.
Claude 5 connects the dots.
'I'm working on [goal] for [who it's for]. They need [what the output enables]. With that in mind: [task].'
2. Context Files
Upload your expertise. Stop explaining in prompts.
"Read these files completely before responding: [filename .md] - [what it contains]."
The file is the brain. This part never changes.
3. Reference
Show Claude 5 what good looks like.
"Reference for what I want to achieve: [paste]."
One example beats ten instructions.
4. Effort
The new change, a few people are talking about.
"This is a [routine / hard / hardest-unsolved] problem. Scope it like it's at the top of your range."
Teams testing Claude 5 on easy tasks undersell it.
Give it your hardest problem.
5. Act
"AskUserQuestion" is still the king.
Add "When you have enough information to act, act. Don't re-litigate my decisions. While weighing a choice, give a recommendation."
6. Scope
Claude 5 over-delivers by default. Control it.
"Do the simplest thing that works well. No extra features, refactors, or abstractions. If I'm describing a problem, the deliverable is your assessment."
The old one did too little. This one does too much.
7. Delegate
One Claude is no longer the limit.
"Split independent subtasks across subagents & keep working while they run. Verify with a fresh-context subagent."
It's not a chatbot anymore. It's a team lead.
8. Evidence
The line that removes fake progress reports.
"Before reporting progress, audit every claim against a tool result. If it's unverified, say so. Tests failed? Show the output."
Anthropic tested this.
It nearly eliminated fabricated status updates.
9. Memory
Claude 5 gets smarter every run. If you let it.
"Record learnings in [notes .md] — one per file. Update, no duplicate. Delete what turns out wrong."
Your prompts expire. Your learning file compounds.
10. Checkpoint
It can run for hours. Decide when it stops.
"Pause only for: destructive actions, scope changes, or input only I can provide. Never end your turn on a promise."
The old fear was Claude stopping too late.
The new fear is stopping too early.
11. Report
The last block. The first thing you read.
"Open with the outcome - the TLDR I'd ask for. Complete sentences. Clear beats short."
It worked for hours. You read for ten seconds.
Copy the full prompt template + download my personal md. files for Claude here:
Step 1. Go to how-to-ai.guide.
Step 2. Subscribe for free. Don't pay anything.
Step 3. Open my welcome email.
Step 4. Hit the automatic reply button inside.
Step 5. Download my .md files. Ready to upload.
I'll fact-check the specific claims here. A few of them are checkable against Anthropic's actual documentation; several are not verifiable and some are shaky.
The first thing worth flagging: there is no publicly available model called "Claude Fable 5" that you can prompt directly. Fable 5 is a Mythos-tier model that isn't generally accessible — and queries to it on certain topics get routed to Opus 4.8 anyway. So a post titled "The Anatomy of a new Claude 'Fable 5' Prompt" is already building on a shaky premise. Let me check whether the underlying technique claims hold up.Let me check the specific "effort" claim and the "evidence/fabricated status updates" claim, since those are the two most concrete factual assertions.I have enough to give you a grounded verdict. Here's the fact-check, section by section, separating what's real from what's fabricated or misleading.
The framing itself is the biggest problem
The post is titled a "Fable 5 Prompt," but Fable 5 isn't a model you prompt directly the way the post implies. It's a Mythos-tier model with safeguards, and there's a whole separate "Prompting Claude Fable 5" doc distinct from the Opus line. More to the point, almost none of the ten "new" techniques are actually Fable-5-specific or new — they're repackaged from Anthropic's long-standing Claude 4.x / Opus prompting guidance. The post takes established advice and dresses it up as insider knowledge about a bleeding-edge model. That's the core misleading move.
What's genuinely accurate
#1 Task (start with why, not what) — Real. Anthropic's docs explicitly say providing context or motivation behind your instructions, such as explaining why the behavior is important, helps Claude better understand your goals and deliver more targeted responses, and that Claude is smart enough to generalize from the explanation. This is legitimate, though it's guidance for the whole 4.x/5 generation, not a Fable 5 novelty.
#3 Reference (show what good looks like) — Real. Few-shot examples are one of Anthropic's most-recommended techniques; examples are described as one of the most reliable ways to steer output format, tone, and structure. The "one example beats ten instructions" line is a fair paraphrase.
#4 Effort — Partly real, but muddled. The effort parameter genuinely exists — it's an API control (low/medium/high, plus xhigh/max on some models) that governs reasoning depth. And it's true that at higher effort on a routine task these models deliberate more than the job needs and can start tidying code you didn't ask them to touch, and that if you see shallow reasoning on a complex task, you should raise the effort level. But the post confuses two different things: the effort parameter (an API setting) versus telling the model in prose "scope it like it's at the top of your range." Those aren't the same lever, and the post treats a written instruction as if it drives the parameter. Also, the actual guidance is nuanced — low and medium often hold quality fine and xhigh is reserved for the hardest work — which is closer to "match effort to the task" than the post's "always give it your hardest problem."
#6 Scope — Real and correctly stated. The fix for over-delivery is a scope instruction, not a lower effort setting, because you usually want the reasoning and not the cleanup. Good advice.
What's shaky or unverifiable
#8 Evidence ("Anthropic tested this. It nearly eliminated fabricated status updates.") — I could not find any Anthropic source supporting the specific claim that an "audit every claim against a tool result" line was tested and "nearly eliminated fabricated status updates." Asking the model to verify claims against tool output is reasonable practice, but the "Anthropic tested this and it nearly eliminated X" framing appears to be an unsupported embellishment. Treat it as marketing, not a documented result.
#5, #7, #9, #10, #11 (Act / Delegate / Memory / Checkpoint / Report) — These are plausible, sensible workflow patterns, and subagent delegation and memory files are real capabilities. But the specific magic phrasings ("Never end your turn on a promise," "one per file, update, no duplicate," etc.) are the author's own conventions, not Anthropic-published wording. They may work fine — they just aren't authoritative or Fable-5-specific.
#2 Context Files ("This part never changes") — The technique is real (CLAUDE.md-style context loading is standard). "Never changes" is overstated given how much prompting guidance shifts between model generations, but it's a minor point.
The call-to-action
The whole thing funnels to "subscribe to my newsletter, hit the auto-reply, download my .md files." That's a lead-generation funnel. Nothing wrong with that per se, but combine it with the fabricated-sounding "Anthropic tested this" claim and the false "Fable 5" novelty framing, and the pattern is: real techniques + invented authority + urgency → email capture. The techniques you can get for free from Anthropic's own docs.
Bottom line
Roughly half the substance is legitimate, well-established prompting advice — mostly #1, #3, #4 (the parameter part), and #6. The problems are the packaging (calling generic 4.x/5 guidance a "Fable 5 secret"), at least one likely-fabricated efficacy claim (#8), and phrasings presented as authoritative that are really just one person's house style. If you want the real version, Anthropic's own prompting-best-practices and effort docs cover it directly and for free.
Given that you run Kactii and write about this stuff — want me to pull the actual Anthropic guidance into a clean, honest template you could use or publish yourself, without the invented claims?