Fable context thrift
Six Claude Code skills that harden Opus 4.8 toward frontier behavior — written by Fable 5, pressure-tested on the target model with transcripts included.
npx -y skills add DizzyMii/fable-skills --skill fable-context-thriftAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Use at the start of any multi-step task and during exploration — before reading files, searching, or re-checking completed work, especially when tempted to read whole files, re-verify known facts, or run independent lookups one at a time.
SKILL.md
2.9 KB, as published. Nobody here has run it
Context Thrift
Overview
Context is the budget everything else is paid from. Spend it on what changes your next action; nothing else.
Rules
- Independent calls go in one parallel block. Multiple greps, reads, status checks — batch them. Sequencing independent reads adds latency, not care.
- Read targeted. The symbol or section, not the file. Widen only when a specific question demands it.
- Delegate sweeps, keep lookups. Broad questions ("where is X handled", find-all-usages in unfamiliar territory) go to a search subagent if available — keep the conclusion, not the file dumps. Known single lookups stay direct. "The subagent is for when I don't know where things live."
- The tool is the confirmation. Never re-read a file to check an edit landed; the edit tool fails loudly. Verify outcomes with one cheap end-state check (a grep for the old token), not per-file re-reads.
- Established facts stay established. What the conversation already verified, don't re-derive. What the user already decided, don't reopen.
- Don't narrate options you won't pursue. Deliberation is for choices you might actually make.
- Enough-to-act test. If more exploration would not change your next action, exploration is over. Act.
Rationalizations
| Thought | Reality |
|---|---|
| "I'll read the whole file for context" | Read what the change touches. The question you're answering defines the lines you need. |
| "Let me just double-check my edit landed" | The tool reported success; silence is success. One end-state grep beats N re-reads. |
| "One more search to be safe" | If the result wouldn't change your action, it isn't safety — it's stalling with receipts. |
| "I'll re-verify the repo layout first" | It's in your context and was correct an hour ago. Directory trees don't rot mid-session. |
| "Careful means one call at a time" | Careful means right calls. Independent lookups in sequence is the same work, slower. |
| "I'll dispatch a subagent to be thorough" | On a 40-file repo, grep returns line-level truth faster than a subagent returns prose. Match the tool to the territory. |
Provenance: mostly predicted — Opus 4.8 baseline-passed this scenario (quotes above are its own passing-run words). fable-skills test logs, 2026-06-10.
Red Flags — stop spending
- A Read call for a file you edited this turn
- The same fact verified twice in one session
- Sequential tool calls with no data dependency between them
- An exploration step you can't name the decision for
- Whole-file reads where a grep already gave you the line numbers