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/yarlson/yarstack/system-investigatenpx skills add yarlson/yarstack --skill system-investigategit clone --depth 1 https://github.com/yarlson/yarstackWrote 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/yarlson/yarstack/system-investigate)<a href="https://agentmods.dev/skills/yarlson/yarstack/system-investigate"><img src="https://agentmods.dev/badge/skills/yarlson/yarstack/system-investigate.svg" alt="Measured on agentmods" 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 | $0.00066 | $0.01349 |
| Opus 5 | $0.00033 | $0.00674 |
| Sonnet 5 | $0.00013 | $0.00270 |
| Haiku 4.5 | $0.00007 | $0.00135 |
Grade A, and why
system-investigate 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 4d 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 — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.
System Investigation
Own one investigation from the user's question to an evidence-backed verdict and practical next step. Treat the prompt as a hypothesis, repository and runtime evidence as authority, and implementation as out of scope.
Establish the investigation
- Extract the question, suspected behavior, affected system, decision it informs, and requested terminal result.
- Write one explicit investigation goal and stopping condition. Use a persisted goal when the host supports it; mark it complete only after the evidence answers the question or proves a precise blocker.
- State the working hypothesis without treating it as fact. Record plausible alternatives that would materially change the conclusion.
- Keep the investigation read-only: do not edit code or durable configuration, commit, push, open or modify pull requests, install dependencies, or mutate live systems. Route a later request to act on the recommendation through the applicable planning or implementation workflow.
- Read all applicable repository instructions before repository-specific work.
Maintain autonomous progress
- Continue without routine confirmation. Resolve discoverable facts before asking the user.
- Check safe sources in this order as applicable:
- repository instructions, source, tests, documentation, configuration, and relevant history;
- installed command help, local versions, APIs, connectors, and authorized browser sessions;
- exact or bounded runtime and remote state;
- official documentation, specifications, upstream source, and release notes for current public facts.
- Never send private issue, repository, runtime, or customer data to public search. Search only the public technical terms needed to answer the question.
- Choose the smallest reversible recommendation when several approaches satisfy the evidence. Record assumptions and proceed.
- Ask one concise question only when the unresolved answer changes required behavior or scope, authoritative sources conflict, necessary access is unavailable, or every safe path requires a high-risk or irreversible action.
- Keep the decision to ask with the owning agent. A subagent may report missing evidence but must not interrupt the user.
- Report inconclusive evidence as inconclusive. Silence and missing tooling are not proof.
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.
- 4d ago First seen · 79 lines · 66 tokens per session scan A bbd2fbeb6ae3
system-investigate is a skill published in the GitHub repository yarlson/yarstack (3 stars, last pushed 7d ago), licensed MIT. It adds 66 tokens to every session and 1,349 once invoked, about $0.0003 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
opensrc
Fetch dependency source code to give AI agents deeper implementation context. Use when the agent needs to understand how a library works internally, read source code for a package, fetch implementation details for a dependency, or explore how an npm/PyPI/crates.io package is built. Triggers include "fetch source for"…
release-cli
Release a new version of the tokenmaxxing CLI to npm. Use when explicitly asked to release or publish @851-labs/tokenmaxxing, including checking main, bumping apps/cli/package.json, updating CHANGELOG.md, committing, tagging cli-vX.Y.Z, pushing, monitoring generated native package publishing, and smoke testing the…
audit-plan-implementation
Independently audit a task attempt or completed run against a frozen plan, checking ownership, evidence, acceptance, and final readiness.
agent-first-planning
Create or revise production-ready SDD plans for agent execution, including clarification, specification, freeze inputs, ownership, budgets, and evidence contracts.
ship-mobile-app
Build, change, debug, and prepare production mobile apps across domain meaning, local and server state, lifecycle, platform, and release boundaries. Use for non-trivial Flutter, React Native, iOS, or Android work involving persistence, sync, auth, offline behavior, notifications, permissions, native integrations…
verify-regression-tests
Verify that a regression test actually detects the defect it claims to guard against. Use after adding or reviewing a bug-fix guard, reproducing the original failure after a fix, or investigating a suspiciously green regression test. Do not use for general TDD, broad test-suite audits, mutation-score optimization…