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 tmusser/ai-engineering-skills --skill lean-modegit clone --depth 1 https://github.com/tmusser/ai-engineering-skillsWrote 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/tmusser/ai-engineering-skills/lean-mode)<a href="https://agentmods.dev/skills/tmusser/ai-engineering-skills/lean-mode"><img src="https://agentmods.dev/badge/skills/tmusser/ai-engineering-skills/lean-mode/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/tmusser/ai-engineering-skills/lean-mode"><img src="https://agentmods.dev/badge/skills/tmusser/ai-engineering-skills/lean-mode.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 | $0.00029 | $0.00860 |
| Opus 5.5 | $0.00012 | $0.00344 |
| Sonnet 5.5 | $0.00006 | $0.00172 |
| Haiku 4.5 | $0.00003 | $0.00086 |
Grade A, and why
lean-mode 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 20d 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 — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Lean Mode
Purpose
Reduce token usage without losing operational clarity.
Compress prose, not meaning.
When to use
Use when:
- the user says "lean mode", "save tokens", "be terse", "compress", "short replies", or invokes
/lean-mode - the task is execution-oriented
- the user already understands the domain
- the conversation is long and token budget matters
- the agent is giving routine status, next steps, commands, or implementation guidance
Do not use when:
- the user asks for teaching or explanation
- the user asks for nuanced trade-off analysis
- ambiguity is high
- safety, security, privacy, legal, medical, or financial reasoning is needed
- the user asks for polished writing
- compression would hide uncertainty or risk
Persistence
Once enabled, stay in lean mode for routine replies until the user says:
- "stop lean mode"
- "normal mode"
- "full reasoning"
- "explain more"
Switch temporarily to fuller prose when correctness requires nuance.
An operator profile changes the default communication density, not the information
boundary. Communication density: lean must never suppress risk, uncertainty,
verification evidence, or material tradeoffs.
Inputs
- User request
- Current task
- Recent context
- Commands, file paths, assumptions, risks, and verification details
- Optional operator profile
Communication density
Workflow
- Detect a lean-mode trigger, explicit request, or an available operator profile
with
Communication density: lean. - If the current request explicitly asks for another level of detail, follow the current request instead of the profile default.
- Compress routine response text.
- Keep exact commands, file paths, assumptions, risks, verification, and next actions.
- Switch back to fuller prose when nuance or correctness requires it.
- Stay terse until the user exits lean mode or a stronger current instruction overrides the profile default.
Style
- Short sentences.
- Bullets over paragraphs.
- No pleasantries.
- No motivational recap.
- No redundant caveats.
- Use arrows for cause/effect.
- Use compact labels:
Verdict,Do,Avoid,Command,Risk,Next. - Use abbreviations only when obvious: repo, config, impl, req, res, fn, env, auth, DB.
- Keep exact filenames, commands, paths, and numbers verbatim.
- Keep uncertainty explicit.
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.
- 20d ago Changed · +9 lines 98499a82695d
- 1mo ago First seen · 161 lines · 29 tokens per session scan A 3ca537b5ef8b
lean-mode is a skill published in the GitHub repository tmusser/ai-engineering-skills (6 stars, last pushed 5d ago), licensed MIT. It adds 29 tokens to every session and 860 once invoked, about $0.0001 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.
Other skills, from other repositories
awsl
Run Claude Code JavaScript Workflows through the awsl compatibility runtime. Use when an agent's current task or loaded Skill requires dispatching a Claude Code Workflow but the host cannot execute that Workflow natively, or when awsl workflow inspection, durable run state, resume, or provider diagnostics are needed.
dwi-all-in-one
Apply the relevant Dwi lenses together when several observed workflow problems co-occur. Select only the lenses the task needs, preserve a silent fast path for clear reversible work, and keep authority and evidence explicit. Prefer a focused module when one issue dominates.
dwi-arc
Structure genuinely multi-agent coding work into bounded cells with one writer per scope, explicit integration, and independent review. Use when several disjoint workstreams justify coordination. Do not use for small tasks, overlapping writers, speculative agent fleets, or process artifacts without demonstrated value.
dwi-bridge
Coordinate bounded work between native Claude and Codex workflows with explicit authority, scope, and evidence. Use for read-only consultation or explicitly authorized execution delegation. Do not create a new connector, share secrets, treat messages as authorization, or allow recursive delegation.
dwi-budget
Set and report practical token, context, time, tool-call, and coordination boundaries for coding-agent work. Use when resource use is unclear or needs a checkpoint. Do not invent measurements, monetary savings, cache benefit, or precision that the harness does not expose.
dwi-evidence
Label coding-agent claims by evidence status, preserve provenance and failures, and separate static, runtime, and human proof. Use before completion, comparison, promotion, or handoff. Do not upgrade observations into guarantees or fabricate missing measurements and approvals.