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/xcrft/mastermind/mastermind-security-researchnpx skills add xcrft/mastermind --skill mastermind-security-researchgit clone --depth 1 https://github.com/xcrft/mastermindWrote 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/xcrft/mastermind/mastermind-security-research)<a href="https://agentmods.dev/skills/xcrft/mastermind/mastermind-security-research"><img src="https://agentmods.dev/badge/skills/xcrft/mastermind/mastermind-security-research.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 | $0.00070 | $0.01105 |
| Opus 5 | $0.00035 | $0.00553 |
| Sonnet 5 | $0.00014 | $0.00221 |
| Haiku 4.5 | $0.00007 | $0.00111 |
Grade A, and why
mastermind-security-research 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 — 113 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Mastermind security research
[[mastermind-agent-security-review]] traces untrusted input to sinks and judges the design. This runs first and answers a narrower question: what can reach the privileged operation, what claims to enforce it, and which of those claims the graph can actually establish.
The output is a population and a gap list. It is not a security verdict, and it must not read like one.
The graph is wrong in both directions here
This is the rule the whole skill hangs on, and security is where getting it wrong costs the most.
A missing edge is not an absence of access. Middleware applied globally, a route table, a decorator-based dispatcher, a DI container, or a proxy in front of the service all invoke code the graph never links. A handler with no static caller may be reachable from the internet.
A present edge is not enforcement. mmcg_callers authGuard returning a
handler proves the name appears in that function. It does not prove the guard
runs before the sensitive operation, that its result is checked, that it fails
closed, or that a branch does not skip it. Order, control flow, and error
handling are invisible to a call edge.
So the graph never answers "is this protected". It produces two lists, and the difference between them is what a human has to read. Say that explicitly in the output — a research packet that lets a reader infer safety from a query result has done harm, not work.
Enumerate what can reach the privileged operation
mmcg_callers <privileged fn> # direct callers
mmcg_impact <privileged fn> --depth 3 # transitive reach
mmcg_api_surface src/admin/ # what outside code actually reaches
Start from the operation that matters — the deletion, the payment, the role change, the token mint — not from the endpoint. The reachable set is the population you must account for; every member is either shown to be enforced or listed as unestablished.
mmcg_api_surface is the empirical attack surface: symbols under a prefix that
code outside it reaches, independent of what is declared public. A module can
export little and still be reached through a re-export.
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.
- 4d ago First seen · 113 lines · 70 tokens per session scan A 82fa8b09892f
mastermind-security-research is a skill published in the GitHub repository xcrft/mastermind (11 stars, last pushed 4d ago), licensed MIT. It adds 70 tokens to every session and 1,105 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 2026-08-30.
Other skills, from other repositories
roam
Codebase comprehension via roam-code CLI. Use when exploring codebases, planning modifications, debugging failures, assessing PR risk, or checking architecture health. Triggers on: understanding project structure, pre-change safety checks, finding symbols/files, blast radius analysis, affected tests, health scoring…
orchestrate
Tree-aware multi-agent GitHub-issue pool. Conductor manages task-tree from task-splitting-evaluation, executing depth-first per branch while parallelizing roots/orphans. Each worker agent owns its subtree lifecycle. Conductor tracks state (pending/started/in-progress/completed/halted) in GitHub +…
bonsai-ninja
Use bonsai-ninja as compiler-backed structural evidence when mapping a codebase, finding symbols, tracing behavior, inspecting dataflow, debugging across files, reviewing change impact, exporting graph facts, or running SAST.
task-splitting-evaluation
Recursive pre-implementation GitHub task splitting and evaluation flow. Use when the user wants Claude agents to evaluate unhandled tasks, skip already-processed tasks, mark easy leaves with detailed executor-ready comments, split broad tasks into GitHub subtasks, and keep recursing until every leaf is well described…
gh-task
Run a GitHub issue end-to-end in this repository using the committed isolated-worktree workflow. Use when the user invokes $gh-task, asks to run a GitHub task, or provides a GitHub issue number.
stacklit-navigator
Use stacklit.json to navigate codebases without burning tokens on file exploration.