Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/rokyytr2/betterthink/codexnpx skills add RokyYTR2/BetterThink --skill codexgit clone --depth 1 https://github.com/RokyYTR2/BetterThinkWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00074 | $0.00801 |
| Opus 5 | $0.00037 | $0.00400 |
| Sonnet 5 | $0.00015 | $0.00160 |
| Haiku 4.5 | $0.00007 | $0.00080 |
Grade A, and why
betterthink scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
Copies of this mod
1 near-identical copy found in the catalogue:
- betterthink — 95% identical, 6 lines differ
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
BetterThink
Apply this skill at the start of a coding task when correctness and discipline matter more than speed.
Clarify Before Editing
- Drive uncertainty down before writing code.
- Ask focused questions when scope, behavior, stack, naming, or edge cases are materially ambiguous.
- If a small assumption is safe and non-destructive, state it explicitly and proceed.
- Do not silently fill important gaps with guesses. Target: reach high confidence in the requested outcome before editing.
Finish One Slice Before Starting The Next
- Work on one concrete to-do at a time.
- Verify the current slice before moving on: read the diff, run the relevant test or command, or inspect the behavior directly.
- Do not stack partially finished fixes with a plan to circle back later unless the task genuinely requires it. Target: treat "probably done" as not done.
Delegate Only When Codex Allows It
Codex may only use subagents when the user explicitly asks for delegation or parallel agent work. Do not force delegation when the runtime instructions do not allow it. When delegation is explicitly allowed:
- Use
explorerfor scoped codebase questions, searches, and fact-finding. - Use
workerfor isolated implementation or verification tasks with a clear file or module boundary. - Keep the main thread on the critical path; delegate sidecar work, not the immediate blocker.
- Reuse the inherited model by default. Override the model only when there is a concrete task-specific reason.
Keep Project Memory Lean
At the end of a meaningful change, update the existing project memory file if the repository already uses one, such as AGENTS.md or another established top-level guide.
Update it only when one of these happens:
- Add a new dependency
- Add a new script or command entry point
- Create a new top-level folder
- Change build, run, test, or deploy commands
- Introduce a new project convention or pattern worth preserving Rules:
- Prefer the file the project already uses; do not create a new memory file unless the repo already expects one or the user asks.
- Record durable information only: architecture, conventions, commands, gotchas, and non-obvious decisions.
- Skip routine bug fixes, refactors, and task logs.
- Keep the file compact. If it starts growing into long-form documentation, move detail into a supplementary doc and leave one pointer line in the memory file.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 56 lines · 74 tokens per session scan A 899c4a5f4483
betterthink is a skill published in the GitHub repository RokyYTR2/BetterThink (1 stars, last pushed 2mo ago), licensed MIT. It adds 74 tokens to every session and 801 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
crisis-catalyst
Reframe a crisis, constraint, or setback as a forcing function for founder-mode decisions that were previously politically blocked. Based on Chesky's pandemic transformation and Andy Grove's "great companies are defined by their crises." Use when facing a serious challenge and want to use it as fuel, not just survive…
founder-quiz
Take a behavioral self-assessment to discover where you sit on the founder mode spectrum. 15 questions that reveal your actual operating mode — not what you think you do, but what you actually do. Use for honest self-reflection.
org-health
Diagnose organizational health against Brian Chesky's framework. Detects the division → politics → bureaucracy → complacency arc that kills founder-led companies. Use when a founder suspects their org structure is slowing them down.
decision-check
Determine whether a decision is a founder-only move (requires institutional memory, passion, and permission) or a manageable decision that can be delegated. Based on Chesky's framework of what only a founder can do. Use when deciding whether to stay involved in or hand off a decision.
hiring-scorecard
Score an executive or senior hire using Brian Chesky's 'guilty until proven innocent' framework. Generates a reference check script, identifies red flags, and assesses builder vs. manager fit. Use when evaluating a candidate for a senior role.
balzac
Balzac is an AI content platform CLI — create workspaces, manage SEO keywords, generate article suggestions, write articles, publish content across integrations, and monitor Google Search Console performance. Supports workspaces, keywords, suggestions, briefings, articles, competitors, links, integrations, search…