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 skills add Redclawww/savethetokens --skill savethetokensgit clone --depth 1 https://github.com/Redclawww/savethetokensWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/redclawww/savethetokens/savethetokens)<a href="https://agentmods.dev/skills/redclawww/savethetokens/savethetokens"><img src="https://agentmods.dev/badge/skills/redclawww/savethetokens/savethetokens/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/redclawww/savethetokens/savethetokens"><img src="https://agentmods.dev/badge/skills/redclawww/savethetokens/savethetokens.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00069 | $0.02304 |
| Opus 5.5 | $0.00028 | $0.00922 |
| Sonnet 5.5 | $0.00014 | $0.00461 |
| Haiku 4.5 | $0.00007 | $0.00230 |
Grade A, and why
savethetokens 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 2mo 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.
How it starts
The opening of the file, as written. The whole thing — 201 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Context Governor
Optimize context usage with practical, high-impact workflows and scripts.
Non-Negotiable Guardrails
- Keep scope locked to the user request. Do not add extra features, pages, or telemetry unless asked.
- Treat token optimization as a constraint, not the goal. Correctness and security win over token reduction.
- Never claim token savings without before/after measurement on comparable tasks.
- If context-saving actions risk quality loss, keep the extra context and state the tradeoff.
- Never reduce code thoroughness to save tokens. Tests, strict config, safety checks, and error handling are non-negotiable. Save tokens from message verbosity — never from output completeness.
Operating Modes
Lean Mode(default): Use lightweight context hygiene only; do not create new benchmark artifacts.Measurement Mode: Use launch-readiness or A/B telemetry scripts only when user asks for proof/percentages.
Claude Code Message Budget (required)
- Keep progress updates short and phase-based. Do not narrate every file write.
- Do not paste long command output unless user asks. Summarize only key signals.
- Do not repeat the same command without a code/input change; if retried, state the reason once.
- If
/contextshows message growth is unusually high, switch to stricter concise mode:- fewer updates
- shorter summaries
- batch related edits before reporting
- Prefer one concise final summary over long running commentary.
- For benchmark runs, enforce matched behavior on both variants:
- same stop criteria
- same compact policy
- same output style (no extra giant report in one variant only)
Operating Playbook
- Confirm objective and lock scope in one sentence.
- Keep one chat session per task. Start a new session for unrelated work.
- Use
! <command>for direct shell commands when no reasoning is required. - Run
/contextperiodically. Compact around 50% usage instead of waiting for hard limits. - Before
/compactor/clear, create a checkpoint file with next steps and touched files. - Keep top-level docs lean; move deep details to linked
docs/*.md. - Before final output on code tasks, run the quality gates in
docs/QUALITY_GATES.md. - For token-savings claims, run matched A/B using
docs/BENCHMARK_PROTOCOL.md. - For Claude benchmark runs, use
docs/STRICT_BENCHMARK_PROMPT.mdas the session starter.
What ships with it
23 files 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.
- scripts/ab_telemetry.py 20 KB runs code
- scripts/analyze.py 8.9 KB runs code
- scripts/classify_intent.py 6.2 KB runs code
- scripts/claude_md_optimizer.py 16 KB runs code
- scripts/compact_watchdog.py 5.8 KB runs code
- scripts/context_snapshot_diff.py 8.5 KB runs code
- scripts/cost_calculator.py 14 KB runs code
- scripts/estimate_tokens.py 4.1 KB runs code
- scripts/govern.py 28 KB runs code
- scripts/launch_readiness.py 13 KB runs code
- scripts/memory_store.py 6.6 KB runs code
- scripts/metrics_collector.py 18 KB runs code
- scripts/path_filter.py 11 KB runs code
- scripts/prune.py 18 KB runs code
- scripts/quick_ref_generator.py 12 KB runs code
- scripts/relevance_scorer.py 17 KB runs code
- scripts/select_model.py 11 KB runs code
- scripts/session_checkpoint.py 9.7 KB runs code
- scripts/session_hook_generator.py 15 KB runs code
- scripts/skill_selector.py 5.2 KB runs code
- scripts/tiered_context.py 15 KB runs code
- scripts/tool_filter.py 4.8 KB runs code
- scripts/utils.py 1.1 KB runs code
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.
- 2mo ago First seen · 201 lines · 69 tokens per session scan A 51bf6487d586
savethetokens is a skill published in the GitHub repository Redclawww/savethetokens (10 stars, last pushed 9mo ago), licensed MIT. It adds 69 tokens to every session and 2,304 once invoked, about $0.0003 per session on Opus 5.5. 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.
savethetokens compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| savethetokensRedclawww/savethetokens | Skill | 10 | 9mo ago | MIT | A | 69 | Claude Code, Codex |
| imagegenopenai/codex | Skill | 135k | today | Apache-2.0 | A | 113 | Codex |
| playwright-climicrosoft/playwright | Skill | 100k | today | Apache-2.0 | A | 19 | Claude Code |
| browser-automationopenclaw/openclaw | Skill | 395k | today | MIT | A | 36 | Claude Code |
| local-ai-agentsmicrosoft/ai-agents-for-beginners | Skill | 80k | 9d ago | MIT | A | 200 | Claude Code, Codex |
| next-cache-components-adoptionvercel/next.js | Skill | 144k | today | MIT | A | 95 | Claude Code |
| chat-pet-sprite-creationmicrosoft/vscode | Skill | 195k | today | MIT | A | 53 | Claude Code, Codex |