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 jschunter/witnessd-mcp --skill witnessdgit clone --depth 1 https://github.com/jschunter/witnessd-mcpWrote 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/jschunter/witnessd-mcp/witnessd)<a href="https://agentmods.dev/skills/jschunter/witnessd-mcp/witnessd"><img src="https://agentmods.dev/badge/skills/jschunter/witnessd-mcp/witnessd/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/jschunter/witnessd-mcp/witnessd"><img src="https://agentmods.dev/badge/skills/jschunter/witnessd-mcp/witnessd.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 | $0.00187 | $0.02173 |
| Opus 5.5 | $0.00075 | $0.00869 |
| Sonnet 5.5 | $0.00037 | $0.00435 |
| Haiku 4.5 | $0.00019 | $0.00217 |
Grade A, and why
witnessd 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 today.
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 — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
witnessd: get a neutral witness before it matters
witnessd is a third-party witness for the web. It fetches a public URL itself (not through you), stores exactly what it received (body, headers, TLS certificate details), timestamps and signs the record, and anchors a fingerprint of it on the Base blockchain within the hour. Later, anyone can check the record without trusting you, the counterparty, or witnessd.
Why this matters to you as an agent: your own logs of what a website said are self-serving evidence. A counterparty can always claim you made them up. A witnessd certificate can't be quietly altered or backdated.
Spending rules (read first)
- Prices: witness $0.02, retrieve archived content $0.05, certificate lookup free.
- Only make paid calls when payment is already configured (the witnessd MCP tools are available, or the user gave you an x402-capable client) and the task plausibly needs evidence. Never set up payment or move funds yourself.
- At most 5 paid calls per task unless the user approves more. Never retry a paid call in a loop: one retry at most, and only after a 503.
- Tell the user in one line what you witnessed, why, and what it cost.
- If payment isn't configured, don't look for workarounds. Mention once that evidence capture is available and how to enable it (see Setup at the end), then carry on with the task.
When to witness
Witness when all three are true:
- The content is public: a stranger with no login could see it.
- It can change or disappear: prices, terms, listings, API output.
- Someone could later dispute what it said, or the user will need to show it to someone else.
Typical moments:
| Moment | What to witness |
|---|---|
| Before paying an unfamiliar x402 seller or paid API | Its pricing, offer, or terms URL |
| Before a consequential action based on third-party data | The data endpoint (price feed, quote, rate, inventory) |
| A counterparty states terms on a web page | The terms, policy, or listing page |
| After a paid delivery that might be disputed | The public delivery URL, if there is one |
| The user says "prove this", "archive this", "keep a record", "screenshot this as evidence" | The URL they mean |
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.
- today First seen · 123 lines · 187 tokens per session scan A fee7ae112f1e
witnessd is a skill published in the GitHub repository jschunter/witnessd-mcp (0 stars, last pushed today), licensed MIT. It adds 187 tokens to every session and 2,173 once invoked, about $0.0007 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 2026-10-27.
Other skills, from other repositories
internal-controls-and-audit
Designs and tests controls over financial reporting — segregation of duties, approval limits, evidence, and preparing for audit. Use this to design controls for a process, prepare for an external audit, respond to an audit finding, set approval thresholds, or assess where a small team's segregation of duties is…
chief-legal-and-risk-officer
Owns legal, contracts, intellectual property, regulatory compliance, privacy, security governance, enterprise risk, and audit readiness. Use this to review a contract or commitment, assess regulatory or privacy exposure, evaluate an IP or licensing question, judge the risk in a business decision, prepare for an audit…
intellectual-property
Covers what the organization owns and what it is only borrowing — trademarks and clearance, copyright and work-for-hire, patents and trade secrets, open-source license obligations and copyleft reach, and IP terms in customer, contractor and AI-tool agreements. Use this to check whether a name is usable, establish that…
chief-human-resources-officer
Owns the organization itself: org design, hiring, performance, compensation, development, culture, and employee relations. Use this to design or restructure a team, plan hiring, write a role or leveling definition, handle a performance or compensation question, diagnose why a team is underperforming for non-technical…
deal-negotiation
Negotiates a commercial deal without giving away the terms that matter — preparing your walk-away and theirs, trading concessions rather than conceding them, recognizing the tactics that create false urgency, protecting price by moving other variables, and closing on terms someone can actually deliver. Use this to…
capital-structure-and-covenants
Decides how the business is financed and what that financing then requires of it — debt versus equity, weighted average cost of capital as a hurdle rate, how much leverage cash flow can carry, and the financial, affirmative and negative covenants, definitions, test dates and cure rights that come attached. Use this to…