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 caglarbozkurt/mcp-heimdall --skill skillgit clone --depth 1 https://github.com/caglarbozkurt/mcp-heimdallWrote 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/caglarbozkurt/mcp-heimdall/skill)<a href="https://agentmods.dev/skills/caglarbozkurt/mcp-heimdall/skill"><img src="https://agentmods.dev/badge/skills/caglarbozkurt/mcp-heimdall/skill/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/caglarbozkurt/mcp-heimdall/skill"><img src="https://agentmods.dev/badge/skills/caglarbozkurt/mcp-heimdall/skill.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 | $0.00071 | $0.00610 |
| Opus 5.5 | $0.00028 | $0.00244 |
| Sonnet 5.5 | $0.00014 | $0.00122 |
| Haiku 4.5 | $0.00007 | $0.00061 |
Grade A, and why
heimdall 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 1mo 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- heimdall — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Heimdall
Run a security scan on an MCP server and report the risk.
When to use
The user is about to install/connect an MCP server and wants to know if it's safe,
or asks to "audit / scan / vet / check" an MCP, or pasted a tools/list output and
asked whether the tool descriptions look poisoned.
How to run
From the Heimdall project directory:
# local server directory
npx tsx src/cli.ts <path-to-server>
# npm package, PyPI package (pypi:name), GitHub URL, or a tools/list JSON dump
npx tsx src/cli.ts <npm-package | pypi:name | github-url | tools.json>
# machine-readable
npx tsx src/cli.ts <target> --json
# also check the server's dependencies for known CVEs (queries OSV.dev over the network)
npx tsx src/cli.ts <target> --online
If installed globally / as a dependency, use heimdall <target> instead.
Add --online when the user wants supply-chain coverage: it checks declared dependencies
against the OSV.dev advisory database and reports real CVE IDs. It sends only dependency
names + versions (never source); leave it off to stay fully offline.
For a behavioral cross-check, heimdall validate <target> runs the server and calls its
tools to observe what it actually does, then compares that to the static flags (confirmed
/ missed / not-exercised). It executes untrusted code with side effects — only suggest it in
a disposable VM/container, never on the user's real machine.
To analyze injection with full fidelity when tool descriptions can't be read from
source, capture the server's tools/list output to a file and pass --tools <file>.
Interpreting the result
- verdict:
pass/warn/fail. Any hard [GATE] finding forcesfail. - risk score: 0 (clean) – 100 (worst).
- fingerprint: stable hash of the tool surface. Save it; if it changes on a later scan, the MCP silently altered its tools (possible rug-pull).
Summarize the gates and highest-severity findings for the user, cite the evidence
(file:line or tool:name), and give a clear install / don't-install recommendation.
Never recommend installing a server with an unresolved GATE finding.
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.
- 1mo ago First seen · 58 lines · 71 tokens per session scan A 2449c39be61c
heimdall is a skill published in the GitHub repository caglarbozkurt/mcp-heimdall (0 stars, last pushed 3mo ago), licensed MIT. It adds 71 tokens to every session and 610 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.
Other skills, from other repositories
agent-safety
Use when bounding an LLM agent that already runs — scoping its task domain, gating tools to least privilege, defending against prompt injection in untrusted web/email/RAG text, requiring human approval on irreversible actions, capping runtime and cost, or triaging what it already did. NOT building the loop, tools, or…
deepseek-and-destroy
Continuously execute a complex multi-phase implementation plan through configurable worker-agent loops until the plan is complete or a genuinely human-level blocker is reached. Uses fresh implementation and review contexts, reviewer-led repair, fresh re-review, durable multi-orchestrator state, and main-orchestrator…
lc-curate-context
Decide which files a task actually needs, record that as a reusable llm-context rule, verify it against the codebase - including the files your selection references but leaves out - and pack it for your own context, a chat, or a sub-agent you dispatch. Load when choosing what code to put in front of a model, packing…
rule8-truth-gate
Use at S4 before an investigation, article, campaign, narration, or screen production with external-world assertions may advance: require returned source bundles, enforce the requested recency interval for every decision-critical premise, reject unmapped items, and emit the mandated PASS/needstool truth record.
evm-audit-flow
Use when you have a Foundry or Hardhat repository of Solidity or Vyper contracts and want the end-to-end EVM review workflow on native CLIs — a slither static pass, semgrep taint rules, mythril symbolic execution, echidna and medusa invariant fuzzing, halmos formal checks, then a forge build-and-test regression.
multi-stance-audit-fanout
Use after automated EVM review has left a small Solidity target needing human-style logic coverage—prepare one source bundle for eight isolated audit stances, then, as the appointed coordinator, deduplicate their returned leads and detect composite chains. Specialists package and consolidate; Chrono alone dispatches…