token-economy

A set of rules for using fewer unnecessary tokens during long coding-agent sessions. Tokens are the text units used to process a request and can affect session cost and limits.

In plain words
What is it for?
Use it during implementation, debugging, or other long tool-using sessions to keep file reads and retries focused.
Why use it?
It avoids repeated file reads and wasteful retries without skipping required context, reasoning, or verification.

Skill for Claude CodeCodex

Part of the fusebase-flow plugin — 56 skills, 2 agents shipped together

Install

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.

agentmods
npx agentmods add skills/fusebase-dev/fusebase-flow/token-economy
Any agent
npx skills add fusebase-dev/fusebase-flow --skill token-economy
Clone the repo
git clone --depth 1 https://github.com/fusebase-dev/fusebase-flow

Made for: Claude Code, Codex.

Or install fusebase-flow, the plugin that ships this one along with the rest of its 56 skills, 2 agents.

Per session 149 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,497 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00149 $0.03497
Opus 5 $0.00075 $0.01749
Sonnet 5 $0.00030 $0.00699
Haiku 4.5 $0.00015 $0.00350

Measured 3d ago against content hash c4606dc4f8c6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

token-economy 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 3d ago.

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.

.agents/skills/token-economy/SKILL.md · 97 lines

How it starts

The opening of the file, as written. The whole thing — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Token Economy (FR-26)

Guardrail first (the rule's first clause)

Quality outranks tokens. These rules eliminate REDUNDANT consumption only — never skip a needed first-read, never thin verification, never truncate reasoning. On any conflict, the correctness/safety floor wins (same floor language as FR-21: ceremony drops, safety never). Citing FR-26 to avoid a read or a verification step inverts the rule.

Rules (each with its quality guard)

Rule Guard / carve-out Canonical home (if pointer)
TE-01 · Read-scoped — read the slice that answers the question (offset/limit, grep→targeted read), not the whole file for one fact Scoped reads are FACT-FINDING; before an EDIT, read enough surrounding context to hold the file's invariants — never grep-and-edit blind
TE-02 · No re-reads of unchanged in-context files — a file already in context is not read again Re-read REQUIRED after invalidation events: your own Edit/Write, hooks/formatters, parallel/delegated agents, git operations, a failed Edit match, context compaction. Uncertain whether it changed → re-read is the CORRECT spend
TE-03 · Generated/vendored restraint — never read large generated, vendored, lock, build, cache, or compiled outputs Unless the generated artifact is itself the subject of the task. Shared definition of generated/vendored = policies/module-size.yml: exempt_globs
TE-04 · Pre-cached identifiers — never re-derive IDs the handoff already carries; verify (one quick read), don't re-discover Stale-looking ID → verify, then surface; don't silently re-derive POINTER → templates/handoff-implement.md § Pre-cached identifiers
TE-05 · Two-strike rule — the same failing approach is not attempted a third time "Same approach" = same action, same inputs, expecting a different result. NOT strikes: FR-10 3/3 reproduction runs, test reruns after a real change, bounded labeled flaky-external retries, a delegate re-dispatch inside the bounded delegate-retry envelope (flow-skills/liveness-discipline) — the approach was never executed. On 2 strikes → diagnose via zoom-out (FR-20) + validation-and-qa (FR-10)
TE-06 · Targeted edits over whole-file rewrites — patch the changed region; don't regenerate the file. This holds for an FR-18 supersede too: supersede replaces stale semantics, not the file — targeted Edit when most sections are unchanged A full Write is the CORRECT primitive (not waste) for structure/mode/ticket changes or when most sections changed — FR-18 mandates the replaced semantics, never the rewrite tool flow-skills/role-discipline/SKILL.md § Supersede Convention
TE-07 · Pointers over reprints / reference-once — in chat and handoffs cite the artifact path, don't repaste its body; once a large body (big tool result, log, JSON, file region) is in context, later turns refer to it by its retrieval handle (path+window, request ID, report path, test name) instead of re-sending it A fresh first inclusion, or a deliberately re-run command's NEW output, is not a re-send; never drop a decision, ID, or evidence the next step needs. Doc-side rule is FR-23 (pointers over restatement) flow-skills/documentation-budget/SKILL.md
TE-08 · Record-then-read — read durable evidence once after the run instead of agent-side polling Exactly two bounded, labeled exceptions: the first live drive of fresh code, and delegate-PROGRESS polling under the bounded delegate-retry envelope (3 attempts / 5 min / one progress read per interval — a delegate that never started leaves no record to read) POINTER → flow-skills/smoke-testing/SKILL.md § Verification cost discipline · flow-skills/liveness-discipline/SKILL.md § Bounded delegate-retry envelope
TE-09 · Delegation economics — delegate only when the sub-agent's context floor costs less than the tokens the split saves Bounded, disjoint slices only flow-skills/task-delegation/SKILL.md

Read the full file on GitHub · 97 lines

Changes

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.

  1. 3d ago First seen · 97 lines · 149 tokens per session scan A c4606dc4f8c6

Subscribe to this mod's changes

token-economy is a skill published in the GitHub repository fusebase-dev/fusebase-flow (9 stars, last pushed 8d ago), licensed MIT. It adds 149 tokens to every session and 3,497 once invoked, about $0.0007 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.

Related

Other skills, from other repositories

lov-finder-action

Generate Mac Finder right-click menu actions. Two modes: (A) Automator Quick Action for file/folder context menus, (B) Finder Sync Extension (Swift + xcodegen) for blank-space context menus. Automatically selects mode based on user intent. Trigger when user mentions "Finder右键", "右键菜单", "Quick Action", "Finder…

lovstudio/skills · 104 tokens

iblai-api-notification

Read and send ibl.ai platform notifications via the API — unread counts, the notifications list with filters, mark-as-read (all, bulk, or specific), delete, the two-step notification builder (preview then send/schedule, with source validation and recipient preview), per-user notification preferences, per-user…

iblai/api · 103 tokens

iblai-api-profile

Read and manage the signed-in user's own profile via the platform API — basic account info, social links, education, experience, résumé, and memory. Use for self-service profile management. (For the user's learning analytics, see /iblai-api-analytics.).

iblai/api · 57 tokens

mischief-managed

Use when finishing a Pi, Codex, Cursor, Claude, or other coding-agent session and wanting to save a lean Markdown session note into your Obsidian vault with decisions, code changes, validation, caveats, and follow-ups.

ethanolivertroy/mischief-managed · 53 tokens

adobe-express-core

Adobe Express environment essentials, two-runtime architecture, project setup, and MCP server configuration. Use when starting add-on development, understanding iframe vs sandbox boundaries, configuring MCP servers, setting up local projects, or validating manifest configuration.

Sandgrouse/adobe-express-dev-skill · 50 tokens

adobe-express-monetization

Monetize Adobe Express add-ons with subscriptions and payments. Use when designing checkout flows, defining subscription tiers, implementing webhook verification, managing entitlements, or securing backend billing logic.

Sandgrouse/adobe-express-dev-skill · 44 tokens